Light microscopy has traditionally relied on lenses to create microscopic images of small objects. Exploiting the recent advances in sensor technologies and digital processing algorithms, lensfree on-chip holographic microscopy takes a different approach to achieve high-resolution optical imaging without compromising the field-of-view. Along these lines, here we present a lensfree optical tomography approach for high-throughput on-chip 3D microscopy applications.
Lensfree Optical Tomography (LOT): 3D Microscopy without the Use of Lenses
We owe much of our knowledge about micro-scale objects, such as cells, to the invention of light microscopy. Together with the advent of three-dimensional (3D) imaging systems, volumetric structural information regarding biological specimen can be routinely obtained using well-established techniques including, but not limited to, confocal microscopy , optical coherence tomography , light-sheet microscopy , optical projection tomography  optical diffraction tomography [5-7] and digital in-line holography . For the same purpose Lensfree Optical Tomography (LOT) has also been recently introduced as a 3D imaging modality for sectional imaging of biological specimens within large sample volumes [9-10]. By replacing lenses and other bulky optical components with digital computation (i.e., reconstruction), LOT is not subject to the limited field-of-view (FOV) and depth-of-field of a typical lens-based optical system. Consequently, it enables 3D imaging of micro-objects over large sample volumes (e.g., >10 mm3). In addition to its enlarged imaging volume, the architectural simplicity of its design also lends itself to a simple and miniaturized platform that is especially suitable for lab-on-a-chip applications. Using only several light-emitting diodes and an optoelectronic sensor array, LOT can offer <1 μm × <1 μm × <3 μm spatial resolution along the x, y and z directions, respectively, over a large sample volume of ~15mm3.
Implementation of LOT
Lensfree optical tomography (LOT) is analogous to X-Ray Computed Tomography (CT) in the way that it measures the transmission of electromagnetic waves through the objects along different directions of illumination.
Imaging & Microscopy Issue 4 , 2012 as free epaper or pdf download
Accordingly, we illuminate the samples, placed directly on the top of an optoelectronic sensor chip, from different directions of illumination (fig. 1a). The light scattered off the cells or micro-organisms interferes with the unscattered portion of the partially-coherent illumination, enabling us to record digital in-line holograms of objects over a large imaging area, without any optical magnification [9-10]. As the wavelength in the optical regime is more than four orders of magnitude longer than hard X-Rays, the recorded patterns are not projection images of the objects as in X-Ray CT, but rather their holographic diffraction patterns. As a result, we first digitally reconstruct these lensfree holograms to undo the diffraction between the object and the sensor planes. To digitally increase the spatial resolution of these reconstructed images beyond that is permitted by the pixel size of the sensor array, we employ pixel super-resolution (PSR) techniques to achieve sub-micron lateral resolution  (fig. 1b).
For weakly scattering objects that are not thicker than the depth-of-focus (e.g., ~40-50 μm after PSR), diffraction within the objects can be ignored, and the reconstructed images can be approximated as projections of a certain property of interest of the objects (such as 3D density function of scattering potential). Eventually, the collection of 2D lensfree projection images obtained at different illumination angles comprises sufficient information to calculate the 3D microscopic images of the objects. To achieve this, we utilize well-established and computationally efficient filtered back-projection algorithms, which involve inverse Radon transformation of high-pass filtered projection images. Once the 3D object function is calculated, slices through the object can be plotted with <1 μm lateral and <3 μm axial resolution to obtain volumetric structural information about the objects. As a result of eliminating the use of lenses, this decent imaging performance can be obtained over a wide FOV of ~15 mm2 and a depth-of-field of ~1 mm, enabling LOT to probe an imaging volume of ~15 mm3.
Depth Sectioning of Large Sample Volumes on a Chip
As a proof-of-concept demonstration, we initially performed tomographic imaging experiments with micro-beads. We placed a chamber having a thickness of ~50 μm directly on a chip, e.g. a CMOS sensor array. By digitally processing the lensfree holograms of these microbeads (spheres with 5 μm diameter) that are randomly distributed within the chamber, we showed that LOT is capable of successful depth sectioning (fig. 2). Although figure 2 shows slice images for a limited region of interest, tomograms can be computed for a large imaging area, as the volumetric data for the entire FOV is obtained with a single data acquisition step. As LOT does not utilize any lenses (such as microscope objectives), depth-of-field (DOF) is practically limited by our detection signal-to-noise ratio (SNR) rather than the numerical aperture (NA) of a lens. Therefore, holograms can typically be recorded for micro-objects that are as far as ~4-5 mm away from the sensor surface. Since the diffracted field at the sensor plane can be digitally back-propagated to the object plane, projection images for any depth-of-interest can be digitally obtained, as long as a sufficient SNR can be maintained. These projection images will have a depth-of-focus of ~40-50 μm (which practically is also object dependant), and using filtered back-projection techniques lensfree tomograms of the objects around this depth region can be rapidly computed (fig. 3). This way, an extended DOF of ~4-5 mm can be sectioned, which enables high-throughput imaging of micro-objects distributed within large sample volumes.
To evaluate the potential use of LOT in life sciences, we also demonstrated slicing of a C. Elegans worm. Figure 4 shows distinct details at different sections through this worm, which provides information otherwise unattainable with 2D lensfree microscopy. These results suggest that LOT can be a powerful tool for high-throughput 3D microscopy applications, particularly for use in lab-on-a-chip platforms.
Dr. Aydogan Ozcan acknowledges the support of the National Science Foundation (NSF) (CAREER Award on BioPhotonics), Office of Naval Research (Young Investigator Award), and the National Institutes of Health (NIH) Director's New Innovator Award, DP2OD006427, from the Office of the Director, NIH. The authors also acknowledge the support of the Gates Foundation, Vodafone Americas Foundation, and the NSF BISH program (under awards 0754880 and 0930501).
 Pawley J. B. (ed.), "Handbook of Biological Confocal Microscopy" Plenum, New York, 1995
 Fujimoto J.G.: Nat. Biotech., vol.21, 1361-1367 (2003)
 Huisken J. et al.: Science, vol. 305, 1007-1009 (2004)
 Sharpe J. et al.: Science, vol. 296, 541-545 (2002)
 Sung Y. et al.: Opt Exp., vol. 17, 266-277 (2009)
 Debailleul M. et al.: Meas Sci Technol, vol. 19, 074009 (2008)
 Charrière F. et al.: Opt Exp., vol. 14, 7005-7013 (2006)
 Xu W. et al.: Proc. Nat. Acad. Sci., vol. 98, 11301-11305 (2001)
 Isikman S.O. et al.: Proc. Nat. Acad. Sci., vol.18, 7296-7301 (2011)
 Isikman S.O. et al.: Lab Chip, vol. 11, 2222-2230 (2011)
Electrical Engineering Department
90095-1600 Los Angeles