Ultrasound pulse guided digital phase conjugation has emerged to realize fluorescence imaging inside random scattering media. Its major limitation is the slow imaging speed, as a new wavefront needs to be measured for each voxel. Therefore 3D or even 2D imaging can be time consuming. For practical applications on biological systems, we need to accelerate the imaging process by orders of magnitude. Here we propose and experimentally demonstrate a parallel wavefront measurement scheme towards such a goal. Multiple focused ultrasound pulses of different carrier frequencies can be simultaneously launched inside a scattering medium. Heterodyne interferometry is used to measure all of the wavefronts originating from every sound focus in parallel. We use these wavefronts in sequence to rapidly excite fluorescence at all the voxels defined by the focused ultrasound pulses. In this report, we employed a commercially available sound transducer to generate two sound foci in parallel, doubled the wavefront measurement speed, and reduced the mechanical scanning steps of the sound transducer to half.
©2012 Optical Society of America
Fluorescence microscopy is an indispensable tool in biology and life sciences [1–3], but unfortunately its imaging depth in biological tissues is rather limited [3–5]. The ballistic component of light, commonly used to form a focus in an imaging system, experiences exponential attenuation with increasing imaging depth due to random scattering. Therefore even techniques such as two-photon microscopy [6, 7] or optical coherence tomography (OCT) [4, 8, 9], known for their superior penetration capability, are restricted to ~1 mm depth in biological tissues. Progress has been reported in characterizing and controlling light scattering in random media [3, 10–25], but practical biological fluorescence microscopy beyond the ballistic regime has been elusive.
Recently, a promising hybrid imaging technique [26, 27] has emerged that combines the molecular sensitivity of light waves with the much deeper penetration of sound waves. An ultrasound focus is used as a virtual guide star that can be freely positioned deep inside highly scattering media. Diffused light that propagates through the ultrasound focus is encoded by a frequency shift. The frequency shifted light is recorded and phase conjugated to form an optical focus at the original position of the ultrasound focus. Combining digital optical phase conjugation (DOPC)  and focused pulsed ultrasound modulation has very recently enabled deep tissue fluorescence imaging with NIR  and visible excitation . In particular, the NIR system is compatible with biological imaging as the Ti:sapphire wavelength is within the low absorption and low photodamage therapeutic window of biological tissues. Moreover, the NIR wavelength inherently penetrates deeper (longer mean free path length) than visible light and promises for 3D two-photon excitation imaging.
Unfortunately, early implementations [28, 29] of this technique have been far too slow for live sample imaging. To form a fluorescence image, the ultrasound transducer needs to be raster scanned inside the sample. At each voxel defined by the focused sound pulse, a new wavefront needs to be measured prior to the fluorescence excitation. As only a tiny portion of the multiply scattered light is sound modulated, a large number of interferograms are required to extract the wavefront at a descent signal to noise ratio (SNR). In the proof of principle demonstration , the wavefront recording per voxel takes 1.2-2.4s, which drastically slows down the overall image acquisition rate. Hence for practical in vivo deep tissue imaging, we need to increase the wavefront measurement speed by orders of magnitude.
Here we present a parallel modulation and measurement scheme to greatly accelerate the wavefront measurements. To this end, multiple acoustic foci at different carrier frequencies serve as distinct virtual guide stars. Since the amount of the sound modulated light is very small compared to the amount of the unmodulated background, the noise floor and hence the SNR for each wavefront measurement remains the same upon parallelization. Therefore, we achieve higher measurement speed without sacrificing the SNR, similar to the advantage of Fourier domain OCT over time domain OCT [8, 9]. As a proof of concept, we generate two acoustic foci with their sound frequencies differing by 2 Hz. Modulated light emerging from both acoustic foci is detected by a Fourier transform of the beating between the sound modulated light and an external reference beam. From the analysis of the sound modulated light in frequency domain, we compared the SNR for parallelized and non-parallelized operation and verified that the SNR remained the same. Using such a scheme, we doubled the wavefront measurement speed and reduced the mechanical scanning steps of the sound transducer by half.
2. Material and methods
2.1 Parallel wavefront acquisition
Figure 1 illustrates schematically the parallel wavefront recording procedure. Two ultrasound pulses with different carrier frequencies are focused simultaneously inside a highly scattering media (Fig. 1(a)). Only a small portion of light propagates through the sound foci and becomes modulated by the focused sound pulses. For the wavefront measurement, we interfere the multiply scattered light with a reference beam. To extract the weak sound modulated light, we acquire a series of interferograms and perform Fourier transform along the time axis. The 2D phase profiles at the corresponding beating frequencies are the wavefronts we use in the subsequent digital phase conjugation. For fluorescence measurements, we phase conjugate the acquired wavefronts in sequence (Figs. 1(b),1(c)) and measure the fluorescence emission power. In such a scheme, we not only double the wavefront measurement speed, but also eliminate the mechanical scanning of the transducer between the two foci.
2.2 Experimental setup
A schematic drawing of the experimental setup is shown in Fig. 2 . Pulsed laser light from a Q-switched Ti:sapphire laser (center wavelength: 778 nm, pulse duration: 20 ns, repetition rate: 10 kHz) is split into a reference beam and a sample beam. The sample beam is frequency shifted by an acousto-optic modulator (AOM) by 50,005,000 Hz and is focused into the sample chamber. An ultrasound transducer (focal length: 0.23 inch, element diameter: 0.25 inch) is mounted on a 3D motorized stage below the sample chamber. Ultrasound pulses are focused inside the sample and are tightly synchronized with the laser pulses. The sample beam, after passing through the sample chamber, is brought to interference with the reference beam on a CMOS camera (Photonfocus, MV1-D2080-160). The CMOS camera and a spatial light modulator (SLM) form the DOPC system . Both devices have the same pixel size and form mirror images through a beam splitter. For fluorescence excitation, the reference beam is used for phase conjugation. The resulting fluorescence emission is filtered by a bandpass filter and imaged onto an EMCCD camera. The camera is not used to form a widefield image of the fluorescence emission but to measure the fluorescence power by hardware binning and summing all remaining large pixels. The typical exposure and camera readout time are 5 and 18 ms, respectively.
2.3 Generation of two ultrasound foci
In our experiments we used a single ultrasound transducer. To generate two ultrasound foci, we launched two focused ultrasound pulses with a short time delay between them. A laser pulse arrived at the moment when the first ultrasound pulse had just passed the center position of the transducer’s focus (as shown in Fig. 3(a) ). The spatial separation between the two pulses corresponded to 50 microns, which is within the Rayleigh range of the transducer’s focus. Figure 3(b) shows the driving signal for the ultrasound transducer. Two truncated sinusoidal signals (carrier frequencies: 50,005,016 Hz and 50,005,018 Hz) of 23 ns duration were combined with a 33 ns delay, which was realized by a customized high speed circuit employing a fast analog switch (Texas Instruments 74LVC1G3157). The signal was amplified to 120 Vp-p before entering the ultrasound transducer. The total electrical power consumption of the transducer is estimated to be ~3.3 mW.
2.4 Timing and synchronization
Figure 4 illustrates the timing and synchronization of our system. The delay generator DG1 served as the master clock, outputting a 10 MHz TTL pulse train to synchronize two arbitrary waveform generators (AWG1 and 2) and the other delay generator DG2. AWG1 output a 50,005,000 Hz sinusoidal signal to drive the AOM. AWG2 provided two CW sinusoidal signals (frequencies 50,005,016 Hz and 50,005,018 Hz). These two signals were truncated and combined by the gating circuit (Gate) to produce the driving signal shown in Fig. 3(b). The gating (TTL pulse) was provided by DG1. In addition, DG1 output two 10 kHz TTL pulse trains to trigger the Q-switched laser and DG2 that controlled the exposure of the CMOS camera. The frame rate of the CMOS camera was 40 Hz, sufficient to sample the beating (16 and 18 Hz, see Fig. 5 ) between the sound modulated light and the reference beam.
2.5 Tissue phantom preparation
For the imaging experiments, we made a fluorescence structure that was completely embedded inside a highly scattering medium. Tissue phantoms were produced by mixing a 1.0 micron diameter polystyrene bead suspension with Agar. The volume ratio was 3:37, which results in a scattering coefficient of 7.09 /mm and an anisotropy factor of 0.9013, as reported in a previous publication . Three holes, arranged in an L-shaped pattern, were manually created on the top surface of a 2 mm thick tissue phantom using a micropipette. The holes had a diameter of ~60 microns and a center to center distance of ~100 microns. We injected 6 micron diameter fluorescence microspheres into the holes and removed the residual beads outside the holes via rinsing. In Fig. 6(a) , a direct widefield image of the fluorescent structure is shown. Subsequently we added a second 2 mm thick tissue phantom on top. As a result, the fluorescent structure was completely embedded inside the tissue phantom. Besides the tissue phantom, the dense fluorescence beads caused additional strong scattering. A direct widefield image through the tissue phantom is shown in Fig. 6(d). Due to random scattering, only a featureless fluorescent blob is visible.
We imaged the fluorescence structure embedded inside the tissue phantom using ultrasound pulse guided DOPC with one (carrier frequency: 50,005,016 Hz) and two acoustic foci. In both modes we acquired 120 interferograms to extract the wavefront. To maximize the SNR, the sample and reference beam were adjusted to have equal power and to nearly fill the full well charge capacity of the camera. At each scan position, we measured the fluorescence signal with DOPC and subtracted the fluorescence background. The background signal was measured by shifting the wavefront by 30 pixels on the SLM, which turned the DOPC ineffective [13, 14, 18, 28]. The scanning step size was 25 microns.
To compare the two operation modes quantitatively, we computed the SNR of the wavefront measurements. Figures 5(a) and 5(b) show the power spectra from the Fourier transform of a temporal series of interferograms with two and one acoustic foci, respectively. The power spectra were averaged over all pixels and the SNR was computed through dividing the peak value by the noise floor (averaged power around the peaks, measured from 10 to 20 Hz). Obviously, the noise floor remained at the same level when the second acoustic focus was added (see Fig. 5). The peak heights remained invariant too, as both foci obtained the same acoustic power. The SNR were 5.4 ± 0.2 and 5.1 ± 0.3 for two and one focus, respectively.
A comparison between the raw image data obtained using two ultrasound foci (Fig. 6(e)) and using only one (Fig. 6(g)) shows only minor differences. We resampled the raw data (e, g) using bicubic interpolation to the same pixel size of the direct widefield images (0.88 microns). The results for the parallelized (Fig. 6(f)) and the non-parallelized mode (Fig. 6(h)) show small differences in brightness and in resolving the individual holes. For comparison, we convolved the direct widefield image (Fig. 6(a)) with a Gaussian bell with a full width half maximum (FWHM) of 38 microns, which corresponds to the PSF of our system . The resulting image (Fig. 6(b)) was resampled to the same pixel size (25 microns) of the DOPC images, as shown in Fig. 6(c).
We propose and demonstrate a scheme to parallelize the detection of multiple wavefronts in ultrasound pulse guided DOPC. Although we have only used two ultrasound foci in this work, it is straightforward to extend the method to N foci. Thereby the time to measure the corresponding wavefronts can be N-times shorter while maintaining the same SNR. This is possible because each sound-modulated wavefront only contributes a very small signal to the full well charge capacity of the camera. Therefore the noise floor remains invariant and since each acoustic focus has the same strength, the SNR is not changed. With such a scheme, the photon budget is greatly improved. In the previous demonstration , light is randomly diffused inside a large volume due to scattering. Only a tiny portion of light propagates through the single sound focus and contributes to the wavefront measurement. The remaining light only contributes to the noise floor. With the parallel modulation scheme, more light is sound modulated (N fold improvement) and therefore the light is used more efficiently. This is analogous to the SNR advantage of Fourier domain OCT over time domain OCT [8, 9]. Moreover, we reduce the mechanical scanning of the sound transducer by a factor of N, further accelerating the image acquisition.
In our experiments we could only place acoustic foci along the propagation direction of the sound wave, since we employed only one ultrasound transducer. Using 2D ultrasound transducer arrays would enable us to place acoustic foci in a 3D volume, which in turn would drastically accelerate 3D image acquisition. If the CMOS camera records M frames of interferograms during one measurement, the number of wavefronts N that can be measured in parallel must satisfy N < M/2 (Nyquist–Shannon sampling theorem). The spatial separation of the acoustic foci can be chosen arbitrarily. In fact they could even overlap, provided that their carrier frequencies are resolvable. This is different to optical projection of multiple foci using monochromatic sources (e.g. Nipkow type spinning disk), where a minimal separation between adjacent foci has to be maintained. This theoretically allows a very dense sampling with a large number of acoustic foci.
In the current implementation, the wavefront recording is the major time-consuming step. Doubling the wavefront recording speed can shorten the overall imaging time to ~half. Employing an array of sound transducers may potentially improve the wavefront recording speed by another factor of ~30. In such cases, the switching time of the liquid crystal SLM (currently ~0.1 sec) needs to be improved to take advantage of the rapidly available wavefront information. Promisingly, large format high speed MEMS based SLMs with millisecond scale switching time are becoming commercially available . By using such a fast SLM and a large area PMT for fluorescence excitation and collection, the overall imaging speed improvement through parallelization will approach ~N.
In summary, we have presented a technique to speed up the necessary wavefront measurements in ultrasound pulse guided DOPC. This is an important step towards the practical application of this technique to live samples. Further speed improvements are expected using cameras with higher full well capacity and higher frame rate. With custom made sound transducer arrays, high speed cameras and SLMs with fast switching time, we expect that the imaging speed can be improved by orders of magnitude. We envision that this will open up exciting possibilities for ultra-deep tissue fluorescence imaging on live samples.
We thank Steven Sawtelle for designing and manufacturing the gating circuit. The research is supported by the Howard Hughes Medical Institute.
References and links
2. B. A. Wilt, L. D. Burns, E. T. Wei Ho, K. K. Ghosh, E. A. Mukamel, and M. J. Schnitzer, “Advances in light microscopy for neuroscience,” Annu. Rev. Neurosci. 32(1), 435–506 (2009). [CrossRef] [PubMed]
3. J. Tang, R. N. Germain, and M. Cui, “Superpenetration optical microscopy by iterative multiphoton adaptive compensation technique,” Proc. Natl. Acad. Sci. U.S.A. 109(22), 8434–8439 (2012). [CrossRef] [PubMed]
4. D. Huang, E. A. Swanson, C. P. Lin, J. S. Schuman, W. G. Stinson, W. Chang, M. R. Hee, T. Flotte, K. Gregory, and C. A. Puliafito, “Optical coherence tomography,” Science 254(5035), 1178–1181 (1991). [CrossRef] [PubMed]
5. P. Theer, M. T. Hasan, and W. Denk, “Two-photon imaging to a depth of 1000 microm in living brains by use of a Ti:Al2O3 regenerative amplifier,” Opt. Lett. 28(12), 1022–1024 (2003). [CrossRef] [PubMed]
8. J. F. de Boer, B. Cense, B. H. Park, M. C. Pierce, G. J. Tearney, and B. E. Bouma, “Improved signal-to-noise ratio in spectral-domain compared with time-domain optical coherence tomography,” Opt. Lett. 28(21), 2067–2069 (2003). [CrossRef] [PubMed]
16. M. Kim, Y. Choi, C. Yoon, W. Choi, J. Kim, Q. H. Park, and W. Choi, “Maximal energy transport through disordered media with the implementation of transmission eigenchannels,” Nat. Photonics 6(9), 583–587 (2012). [CrossRef]
17. M. Cui, E. J. McDowell, and C. H. Yang, “Observation of polarization-gate based reconstruction quality improvement during the process of turbidity suppression by optical phase conjugation,” Appl. Phys. Lett. 95(12), 123702 (2009). [CrossRef] [PubMed]
19. M. Cui and C. Yang, “Implementation of a digital optical phase conjugation system and its application to study the robustness of turbidity suppression by phase conjugation,” Opt. Express 18(4), 3444–3455 (2010). [CrossRef] [PubMed]
21. M. Cui, “A high speed wavefront determination method based on spatial frequency modulations for focusing light through random scattering media,” Opt. Express 19(4), 2989–2995 (2011). [CrossRef] [PubMed]
22. O. Katz, E. Small, Y. Bromberg, and Y. Silberberg, “Focusing and compression of ultrashort pulses through scattering media,” Nat. Photonics 5(6), 372–377 (2011). [CrossRef]
23. D. J. McCabe, A. Tajalli, D. R. Austin, P. Bondareff, I. A. Walmsley, S. Gigan, and B. Chatel, “Spatio-temporal focusing of an ultrafast pulse through a multiply scattering medium,” Nat Commun 2, 447 (2011). [CrossRef] [PubMed]
27. P. X. Lai, X. Xu, H. L. Liu, Y. Suzuki, and L. V. Wang, “Reflection-mode time-reversed ultrasonically encoded optical focusing into turbid media,” J. Biomed. Opt. 16(8), 080505 (2011). [CrossRef] [PubMed]
28. K. Si, R. Fiolka, and M. Cui, “Fluorescence imaging beyond the ballistic regime by ultrasound-pulse-guided digital phase conjugation,” Nat. Photonics 6(10), 657–661 (2012). [CrossRef]
29. Y. M. Wang, B. Judkewitz, C. A. Dimarzio, and C. Yang, “Deep-tissue focal fluorescence imaging with digitally time-reversed ultrasound-encoded light,” Nat Commun 3, 928 (2012). [CrossRef] [PubMed]
30. M. Wildenhain, J. Knobbe, A. Gehner, M. Wagner, and H. Lakner, “Ao slm demonstration system and test bed,” Proc. SPIE 6467, 64670E, 64670E-10 (2007). [CrossRef]