Led by innovations in the use of stimulated emission or image processing, a number of novel microscopes that break the diffraction limit of light have emerged in the last decade [1, 2]. However, the throughput of microscopes, currently limited by image sensor technology, has not received much attention. Yet, when combined with existing microfluidics technology (e.g., flow cytometry), high-throughput imaging microscopy enables label-free screening of large volumes of blood or tissue. Without requiring fluorescent tags, it will be able to identify rare diseased cells among a large population of healthy cells for diagnosis of early stage cancer, spread of cancer to lymph nodes in the vicinity of a primary tumor, and determining the efficacy of drug and radiation treatments. Today, optical microscopy performed on stationary cells that are fixed on a glass slide, the so-called blood smear test, is the standard diagnosis tool in hematology and pathology. Yet, this technique has limited statistical accuracy because it only screens a small volume of blood placed on a glass slide. Indeed, a high-throughput cell screening method will improve the statistical accuracy of pathology tests based on cell imaging. High-throughput microscopy can also address the need for the purification of specific subpopulations of stem cells: undifferentiated cells that differentiate to specific specialized cells.
To be sure, regenerative medicine based on stem cell transplantation is viewed as one of the most important fields in the health sciences . To ensure the fidelity of transplanted cells, combining high-throughput microscopy and laser ablation may hold great promise for ex vivo purging of undesired cells from stem cell grafts. Further, we believe that by merging high-throughput imaging with next-generation laser ablation technology (as described below), noninvasive detection and elimination of aberrant cells in vivo and in situ will become a reality. When automated, this tunable process will seek and destroy such high-value targets as circulating tumor cells (CTCs) [4, 5, 6]. CTCs are diseased cells that are known to be forerunners of metastasis: the spread of cancer from a primary tumor that leads to 90% of mortalities in cancer patients.
High-throughput imaging, however, implies short frame intervals during which few photons are collected, leading to loss of sensitivity. While high-end image intensified CMOS cameras are now able to perform imaging at speeds approaching 100,000 frames per second, they require high-intensity illumination to overcome this fundamental loss of sensitivity at high speeds. Therefore, conventional image sensors (CCD and CMOS based) are not well suited for high-speed microscopy because the high illumination intensity they require can alter or damage the cell especially when focused onto micrometer-scale fields of view. Their relatively long shutter speed (hundreds of nanoseconds) results in blurring and loss of resolution during high-speed flow. This makes them unsuitable for screening a large number of cells in a reasonably short time. Another limitation on frame rate is the time needed to download the image from the array of pixels (readout time). To achieve high frame rates, the number of pixels that are read must be reduced in a process known as the partial readout . This leads to reduced image resolution at high frame rates.
High-Throughput Real-Time Imaging by
Spectral Encoding and Time-Stretch Amplified Fourier Transform
A fresh approach to imaging that has established the record in frame rate and shutter speed for a continuous running camera has recently been demonstrated [8, 9]. This system performs a two-step process. In the first, the spectrum of a broadband optical pulse is stamped with the spatial pattern of the sample, with each laser pulse capturing one frame. In the second step, the image-encoded spectrum is mapped into a time-domain serial stream that is detected, not with a multiple-pixel image sensor array, but with a single-pixel high-speed detector. Concurrently, the stream is optically slowed down so that it can be digitized in real time with an electronic analog-to-digital converter such as a digital oscilloscope. Called serial time-encoded amplified microscopy (STEAM), this new imaging technique continuously captures a real-time movie of a fast dynamic process because it slows the time scale of the process before the image is detected by the photosensor and digitizer. We have previously created and applied this approach to real-time instrumentation for creating ultra wideband analog-to-digital converters [10, 11], real-time spectroscopy [12, 13], barcode reading , and optical coherence tomography (OCT) .
Serial Time-Encoded Amplified
STEAM achieves a frame rate that is nearly two orders of magnitude higher than CMOS and CCD cameras with three orders of magnitude faster shutter speed. The first step is to create a one-to-one mapping between the spatial coordinates of an object and the optical frequencies by illuminating the sample with a 1D or 2D rainbow created with diffractive optics (Figure 1). A single diffraction grating is used in the 1D line scan mode whereas an orthogonally oriented pair of a diffraction grating and a virtually imaged phased array is employed in the 2D mode. Depending on the reflectivity of the object, the instrument can be operated in either reflection or transmission modes. The different frequency components of the pulse are recombined using the same optical device before an optical circulator directs the image-encoded pulse into a novel time-stretch amplified Fourier transformer realized with an optical fiber with high group-velocity (temporal) dispersion and low loss. Here, the image is stretched into a 1D (analog) data stream because each pixel, now represented by a different frequency component of the pulse, experiences a different time delay. At the same time, the image is optically amplified by pumping the dispersive medium to produce optical amplification via stimulated Raman scattering. Resembling the signal in an optical communication link, the amplified spatial image appears as a time-modulated waveform and is captured with an analog-to-digital converter such as a digital oscilloscope. Given the distributed nature of the loss in the dispersive element, Raman amplification in the dispersive medium is preferred to minimize the noise figure, although discrete optical amplifiers can also be used as well. Animated movies that illustrate the operation of 1D and 2D STEAM can be viewed on the internet at sites  and , respectively. The performance of the STEAM imager, in particular, its diffraction-limited spatial resolution, has been modeled and quantified . Because cells are mostly transparent, a phase-contrast mode of STEAM is preferred to avoid having to stain biological samples with chemical reagents. The phase-contrast mode for high-speed 3D imaging has already been demonstrated in our group .
|Figure 1. Serial time-encoded amplified microscopy (STEAM). The spectrum of a broadband optical pulse is spread onto a 2D plane with the spatial disperser. The image of the sample is encoded into the spectrum of the reflected pulse, which is then directed toward the optically pumped dispersive fiber. Here the image-encoded spectrum is serialized into a temporal waveform by the group-velocity dispersion of the fiber and simultaneously amplified in the optical domain. The amplified temporal waveform is detected by the single-pixel photodetector and digitized by an analog-to-digital converter such as a real-time oscilloscope.
|Figure 2. Real-time observation of laser ablation dynamics enabled by STEAM. (a) Experimental arrangement. A mid-infrared pulse with 5 ns pulse width is focused at an angle onto the sample while STEAM pulses are incident normal to it to monitor the dynamics of laser ablation. (b) Real-time STEAM images of the laser ablation dynamics at a frame rate of 6.1 MHz corresponding to a frame repetition period of 163 ns. (c) Evolution of the normalized reflectivity map of the sample. The observation is consistent with the phase explosion effect.
Previously, we have shown that STEAM captures real-time images of fast microfluidic flow at rates as high as several m/s . We note that 5 m/s is currently the maximum flow rate of conventional flow cytometers – systems that have a throughput of roughly 100,000 cells per second. The shutter speed of the STEAM imager is on the order of 100 ps, compared to 100 ns for ultrafast CMOS cameras. This fast shutter speed is crucial for avoiding image blurring caused by the fast moving cell. In microfluidic imaging experiments described here, the optical source was a mode-locked femtosecond fiber laser with the center wavelength 1590 nm, bandwidth of about 15 nm, and pulse repetition rate of 6.1 MHz. We note that operation at shorter wavelength would be preferred to reduce water absorption. The 2D disperser consisted of a diffraction grating with a groove density of 1200 lines/mm and a virtually imaged phased array with a free spectral range of 67 GHz and a linewidth of 550 MHz. These specifications may vary depending on the available optical bandwidth and the desired number of pixels. Furthermore, a net optical image gain of 25 dB was achieved, enabling frame rates as high as 6.1 million frames per second and shutter speed of a few hundred picoseconds. In 1D line scan mode, the system is operating at 37 MHz frame rate with similar shutter speed.
The STEAM imager was also used to observe the dynamics of pulsed laser ablation in real time . In the preliminary experiment described in Figure 2a, a mid-infrared pulse laser was focused at an angle onto a test sample while the incident imaging pulse train was reflected at normal incidence. Figure 2b shows the images of the laser ablation phenomena at a frame rate of 6.1 MHz. STEAM captures, in real time, the frame sequence corresponding to the dynamics of the ablation caused by the single pulse. STEAM shows that the surface reflectivity changed after a finite time delay (Figure 2c). The change is due to mass ejection from the surface and is a signature of the so-called phase-explosion effect that is the hallmark of laser ablation.
Spectral-shower Encoded Confocal
Optical Microscopy and Microsurgery (SECOMM)
A complementary spin off from STEAM known as spectral-shower encoded confocal optical microscopy and microsurgery (SECOMM) takes advantage of the fast real time imaging capability . Illustrated in Figure 3a, it is a new fiber probe that performs ablation and imaging simultaneously while eliminating the need for mechanical scanning of the laser scalpel. In addition, the probe performs confocal microscopy as the aperture of the fiber that captures the reflection from the sample rejects the scattered light from out-of-focus axial planes.
|Figure 3. Spectral-shower encoded confocal optical microscopy and microsurgery (SECOMM). (a) SECOMM system design. Spectrum of the broadband light source is mapped onto a plane via the spatial disperser for imaging the biological tissue. The back-reflected spectrum is detected by the optical spectrum analyzer for image reconstruction. The high-power frequency-tunable laser is combined with the broadband light to perform laser ablation of the tissue. (b) Reflectivity map of a bovine tissue before and after the laser microsurgery. By tuning the frequency of the ablation laser, an L-shaped pattern within an area of 40 µm 3 40 µm on the tissue was ablated.
Laser ablation is a ubiquitous technology with numerous existing applications such as laser surgery, cutting, and micromachining. Enabled by STEAM, high-speed monitoring of laser ablation dynamics is extremely valuable because when combined with a feedback loop, laser power can now be adjusted in real time in such a way to dynamically adjust the power and exposure time to optimize the desired results while minimizing collateral damage. In the future, laser induced cell death may become a powerful tool for purifying populations of adherent cells for regenerative medicine based on stem cell transplantation. In many applications, the maximum value for real-time monitoring and control of laser ablation will be realized when the laser beam can be steered with high precision at high speed, a requirement that points to non-mechanical laser scanning.
To perform simultaneous imaging and in situ laser ablation, a frequency-tunable continuous-wave laser amplified by an erbium-doped fiber post-amplifier is added to STEAM (Figure 3a). The frequency-tunable laser performs the ablation as its frequency, and that of the STEAM imaging pulses, are mapped to the 2D spatial coordinates of a sample by a common diffractive optics arrangement (Figure 1 and Figure 3a). In experiments performed on a bovine tissue sample, the ablation is performed by the 300 mW frequency-tunable laser/amplifier via the laser-induced thermal coagulation due to the strong water absorption of the tissue at around 1550 nm. By tuning the wavelength of the laser, the ablation beam is directed to any arbitrary position on the tissue without any mechanical movement of the probe or the tissue. Hence, high-precision microsurgery can be performed by computer-controlled tuning of the laser wavelength according to a pre-programmed pattern and at speed much greater than manual scanning.
Figure 3b shows the images of the tissue captured by SECOMM before and after the laser ablation . To simplify the first demonstration of the simultaneous imaging and microsurgery, an optical spectrum analyzer was used instead of the time-stretch amplified dispersive Fourier transformer. The random variations appearing in the figure represent the reflectivity profile of the tissue. By tuning the frequency of the ablation laser, an L-shaped pattern within an area of 40 µm x 40 µm on the tissue was ablated.
The proof-of-concept experiment illustrates the capability of SECOMM to perform in situ high-precision laser microsurgery and simultaneously monitor the process with the same single-fiber probe. This capability bodes well for microsurgery applications that require higher precision than what is achievable with manual manipulation of the surgical probe. Some of the microsurgical applications envisioned for this device include cell microdissection and neurosurgery. In both cases, the high precision that is enabled by replacing mechanical steering with electronic frequency tuning will be a valuable asset.
Ultrafast Optical Coherent Tomography (OCT)
A reflectometry method known as OCT [20, 21] provides cross-sectional images of near-surface internal structure of biological tissue, with important applications in dermatology, ophthalmology, and gastroenterology. In OCT, optical reflections from different layers along the depth (axial) dimension are converted to intensity changes using an interferometer. For each wavelength of the light, constructive interference leading to maximum intensity occurs for reflection from a different layer. This process encodes the depth profile of the sample onto the optical spectrum. High-speed OCT is needed to “freeze” tissue motion and reduce motion artifacts . A high axial scan rate is, hence, desired in OCT in which axial line scans are performed at many points on the surface to capture the volumetric image.
|Figure 4. Ultrafast optical coherence tomography (OCT) enabled by amplified time-stretch Fourier transformation. Optical pulses reflected from different layers of the sample interfere with those from the reference arm of the interferometer. Interfered pulses are then subject to amplified time-stretch Fourier transform as the result of which their spectrum is converted into an amplified temporal waveform. This is subsequently detected by the single-pixel photodetector and digitized by the real-time oscilloscope. Digital inverse Fourier transformation is performed on each pulse to obtain the axial profile of the sample that was encoded onto the spectrum by the interferometer.
The application of time-stretch amplified Fourier transformation to OCT has achieved a record scan rate of 37 MHz  – nearly 1000 times faster than conventional high-speed OCT. As shown in Figure 4, the axial profile of the sample is first mapped onto the spectrum of a femtosecond optical pulse using optical interferometry. The time-stretch amplified Fourier transformer then converts the spectrum into a time-domain serial stream into which the axial profile is encoded. At the same time, it amplifies the signal to compensate for the unavoidable loss in the dispersive medium and hence, overcome the thermal noise limit of the photodetection circuitry. Furthermore, by eliminating the need for a traditional CCD-based spectrometer, time-stretch amplified Fourier transformation significantly increases the axial scan rate and achieves high sensitivity without the need for high-power optical sources.
This article has reviewed recent results on a new high throughput imaging technique and its application to microscopy and laser microsurgery. The technology has the potential to find rare cancer cells among a large population of healthy cells, and possibly, to purify the populations via real-time laser induced elimination of those cells. We have described our vision for how this new technology may advance cancer diagnostics and regenerative medicine. We close by adding that, as an immediate application, this technology may find use in industries where real-time monitoring, dynamic control and optimization of manufacturing processes are needed.
The authors would like to thank Dr. Shahrooz Rabizadeh and Dr. Kayvan Niazi of Abraxis BioScience and the California Nanosystems Institute of Abraxis BioScience and the California NanoSystems Institute for reviewing this article and for his valuable suggestions.
Technology development leading to the STEAM imager was supported by the Micro Technology Office (MTO) of the Defense Advanced Research Projects Agency (DARPA).
- S. W. Hell, “Toward fluorescence nanoscopy,” Nature Biotechnology 21, 1347 (2003)
- B. Huang, W. Wang, M. Bates, and X. Zhuang, “Three-dimensional super-resolution imaging by stochastic optical reconstruction microscopy,” Science 319, 810 (2008)
- K. Mossman, “The world’s first neural stem cell transplant,” Scientific American, December 18 (2006)
- Deakin University, “Rogue cells could cause spread of breast cancer,” ScienceDaily, June 14 (2007)
- J. Kaiser, “Cancer’s circulation problem,” Science 327, 1072 (2010)
- S. L. Stott et al., “Isolation and characterization of circulating tumor cells from patients with localized and metastatic prostate cancer,” Science Translational Medicine 2, 1 (2010)
- http://www.shimadzu.com/products/test/hsvc/oh80jt 0000001d6t.html#
- K. Goda, K. K. Tsia, and B. Jalali, “Serial time-encoded amplified imaging for real-time observation of fast dynamic phenomena,” Nature 458, 1145 (2009)
- K. Goda, K. K. Tsia, and B. Jalali, “Amplified dispersive Fourier-transform imaging for ultrafast displacement sensing and barcode reading,” Applied Physics Letters 93, 131109 (2008)
- F. Coppinger, A.S. Bhushan, and B. Jalali, “Time magnification of electrical signals using chirped optical pulses,” Electronics Letters 34, 399 (1998)
- Y. Han and B. Jalali, “Photonic time-stretched analog-to-digital converter: fundamental concepts and practical considerations,” Journal of Lightwave Technology 21, 3085 (2003)
- P. Kelkar, F. Coppinger, A. S. Bhushan, and B. Jalali, “Time-domain optical sensing,” Electronics Letters 35, 1661 (1999)
- J. Chou, Y. Han, and B. Jalali, “Time-wavelength spectroscopy for chemical sensing,” IEEE Photonics Technology Letters 16, 1140 (2004)
- K. Goda, D. R. Solli, and B. Jalali, “Real-time optical reflectometry enabled by amplified dispersive Fourier transformation,” Applied Physics Letters 93, 031106 (2008)
- K. K. Tsia, K. Goda, D. Capewell, and B. Jalali, “Performance of serial time-encoded amplified microscope,” Optics Express 18, 10016 (2010)
- K. Goda, A. Motafakker-Fard, and B. Jalali, “Phase-contrast serial time-encoded amplified microscopy,” CLEO Europe, June 14 (2009)
- K. K. Tsia, K. Goda, D. Capewell, and B. Jalali, “Simultaneous mechanical-scan-free confocal microscopy and laser microsurgery,” Optics Letters 34, 2099 (2009)
- D. Huang et al, “Optical coherence tomography,” Science 254, 1178 (1991)
- D. C. Adler, Y. Chen, R. Huber, J. Schmitt, J. Connolly, and J. G. Fujimoto, “Three-dimensional endomicroscopy using optical coherence tomography,” Nature Photonics 1, 709 (2007)
Bahram Jalali is a Professor of Electrical Engineering at UCLA, a fellow of IEEE and of the Optical Society of America, and recipient of the R. W. Wood Prize from the Optical Society of America. In 2005, he was elected into the Scientific American Top 50, and received the BrideGate 20 Award in 2001 for his contributions to the Southern California economy. Jalali serves on the Board of Trustees of the California Science Center and the Board of Columbia University School of Engineering and Applied Sciences. He has published over 300 journal and conference papers and holds 7 patents.
Keisuke Goda obtained a B.S. summa cum laude from UC Berkeley in 2001 and a Ph.D. from MIT in 2007, both in physics. He is currently a postdoctoral fellow working on optoelectronics and biophotonics in Photonics Laboratory at UCLA. He is the co-chair of IEEE Photonics Society Los Angeles Chapter and recipient of the Gravitational Wave International Committee Thesis Award and UCLA Chancellor’s Award for Postdoctoral Research. He has published over 80 journal and conference papers.
Patrick Soon-Shiong is Executive Chairman, Abraxis BioScience and Executive Director of the UCLA Wireless Health Institute, and Professor of Bioengineering and Microbiology, Immunology, and Molecular Genetics at UCLA. He is a fellow of the American College of Surgeons and the Royal College of Physicians and Surgeons of Canada. He invented the first FDA-approved protein nanoparticle delivery technology for the treatment of metastatic breast cancer and in trials for lung, melanoma, gastric and pancreatic cancer. He holds over 50 U.S. patents and has published more than 100 scientific papers. He serves on numerous boards and his awards include The Ellis Island Medal of Honor, the Partners in Care Foundation’s Champion for Health Award and the Friends of National Library of Medicine’s 2010 Distinguished Medical Science Award.
Kevin K. Tsia
Kevin K. Tsia received B.E. and M.Phil. degrees in electronic and computer engineering from the Hong Kong University of Science and Technology in 2003 and 2005, respectively. He received his Ph.D. from UCLA in electrical engineering in 2009. He is currently an assistant professor in the department of electrical and electronic engineering and the medical engineering program at the University of Hong Kong. He is a recipient of the California NanoSystems Institute fellowship and Harry M. Showman Prize from UCLA.