Initially, to test the performance of our lensfree on-chip imaging platform, we imaged fluorescent micro-beads (4 µm diameter, Excitation: 505nm, Emission: 515nm) by using two different sensor chips as illustrated in . In specific, we worked with two full-frame CCD chips namely, KODAK KAF-8300 (5.4 µm pixel size, ~2.4 cm2 active imaging area) and KODAK KAF-39000 (6.8 µm pixel size, ~18 cm2 active imaging area). In our lensfree fluorescent imaging modality, because the fluorescent detection occurs at extremely oblique angles on the sensor chip, depending on the opto-electronic design of the pixels and the underlying circuitry of a given chip, the fluorescent point-spread function (PSF) of our platform would exhibit a noticeable variance in its 2D pattern from one sensor-chip to another, which requires calibration of each chip by measuring its unique PSF. Therefore the main purpose of using different sensor chips in this work was to demonstrate sensor independent performance of our lensfree imaging modality.
For this end, we first measured the fluorescent PSF of our lensfree platform by imaging isolated 4µm fluorescent beads using these two different CCD chips. and illustrate these measured point-spread-functions for KODAK KAF-8300 and KODAK KAF-39000 CCD chips, respectively. As demonstrated in these figures, the PSF of each sensor, under similar imaging conditions, is quite different from the other, which is mostly dictated by the opto-electronic design of each CCD chip. For instance, the full-width-half-maximum (FWHM) of KAF-8300 PSF is ~80µm, which implies a fairly limited resolving power for raw fluorescent images as illustrated in . The same conclusion also applies to KAF-39000 PSF with an FWHM of ~120µm, as a result of which, closely packed 4µm fluorescent particles cannot be resolved from each other in raw lensfree images (see ).
Lensfree fluorescent images of various 4 µm bead-pairs.
Decoding performance of our lensfree fluorescent imaging platform with a different sensor chip.
On the other hand, compressive decoding of these raw lensfree images (using the measured PSFs) permits close to an order-of-magnitude increase in our resolving power by rapid digital reconstruction of the fluorescent distribution at the object plane (for further details refer to the Experimental Methods Section). The performance of this compressive decoding approach is quantified in and (for KAF-8300 and KAF-39000 chips, respectively), which both indicate a resolution of ~10 µm that is independently confirmed using conventional fluorescent microscope images of the same 4µm particle pairs (refer to the inset images in , ). These experimental results successfully demonstrate the sensor-chip independent decoding performance of our lensfree fluorescent imaging platform.
The resolution limit in our lensfree imaging results is mainly dictated by the detection signal-to-noise-ratio (SNR), since the tails of the measured PSF, after a certain signal strength, fall below the noise floor of the sensor. In these reported experiments (, ) the CCD chips were kept in room temperature, and therefore further improvement in resolution (beyond ~10µm) can potentially be achieved by active cooling of the opto-electronic sensors without a trade-off in the imaging FOV, which spans the entire active area of the CCD, i.e., ~2.4 cm2 for KAF-8300 and ~18 cm2 for KAF-39000 (see ). We should also note that, with larger area sensors, the imaging FOV of this platform can be even further increased while maintaining a similar resolution level.
On a related note, it is important to emphasize that the pixel size in lensfree compressive imaging is “not” a fundamental limitation for spatial resolution if the detection SNR is sufficiently high. Consider for instance lensfree imaging of two fluorescent points that are directly located on a single pixel. Under this condition, it is theoretically and practically impossible to resolve these two fluorescent points that fall within a single dummy pixel. However, the same two sub-pixel fluorescent points can be resolved from each other using lensfree compressive imaging if several pixels could detect weighted averages of their fluorescent emission. Therefore, under an appropriate detection SNR, if the physical gap between the fluorescent objects and the sensor plane can be increased to perform efficient spatial encoding of the fluorescent objects, resolving of arbitrarily sub-pixel point sources would be feasible. The fundamental limitation to this resolving power is therefore the detection SNR, which determines how many pixels can independently and accurately measure the lensfree fluorescent contributions of the particles. Therefore, for a practical SNR level, there is always an optimum gap range between the object and sensor planes, which we found to be ~50–200 µm for our CCD chips at room temperature.
After this initial characterization of the performance of our wide-field fluorescent imaging platform, we next imaged transgenic C. elegans
samples (refer to the Methods
Section for details) over a wide FOV using the same lensfree configuration depicted in . The results of these imaging experiments are summarized in and (as well as Appendix S1
, see e.g., Figures S3
), which also provide conventional fluorescent microscope images of the same samples for comparison purposes. As shown in these figures, raw lensfree fluorescent signatures of the worms are highly blurred due to our broad PSFs. However, using the measured PSF of each platform, these lensfree signatures can be compressively decoded to digitally yield much higher resolution images of the fluorescent regions located within the C. elegans
body, which very well agree with the images obtained using a regular lens-based fluorescent microscope (see , ). These experimental results successfully demonstrate the efficacy of our compressive decoding approach to image transgenic C. elegans
samples using lensfree fluorescent on-chip imaging over an ultra-wide FOV that covers the entire active area of the CCD chip (e.g., >2–8 cm2
Lensfree fluorescent and holographic transmission imaging of C. elegans.
Lensfree imaging of transgenic C. elegans with a different sensor chip.
We should also note that, the presented on-chip microscopy platform could potentially achieve multi-color imaging of biological samples labeled with multiple distinct targets. In our reported experiments, monochrome CCD chips were used to achieve single color lensfree fluorescent imaging; however, the use of e.g., RGB CCD chips could be utilized to image multiple colors. Unlike conventional lens-based fluorescent microscopy the use of an RGB sensor chip does not immediately bring color imaging capability since without the use of any lenses, all the colors mix with each other at the sensor plane due to unavoidable diffraction. Therefore, lensfree fluorescent imaging might require a more sophisticated compressive decoder to enable separation of multiple colors using raw format RGB images, which was not at the focus of this work.
In addition to fluorescent imaging, our lensfree on-chip platform also permits holographic transmission imaging 
of the worms using the top interface of the same prism that is used in fluorescent excitation (see ). In this lensfree holographic imaging approach, a spatially incoherent quasi-monochromatic source such as a light-emitting-diode (LED) illuminates the samples of interest after being spatially filtered by a large aperture (e.g., 0.05–0.1 mm diameter). This incoherent light source picks up partial spatial coherence that is sufficiently large to record lensfree in-line holograms of the worms on the CCD chip. These acquired in-line holograms can then be rapidly processed using iterative recovery algorithms 
to create lensfree transmission images of the C. elegans
samples over the entire active area of the sensor-chip, matching the imaging FOV of the fluorescent channel. , illustrate such reconstructed lensfree holographic images of the samples, where the lensfree fluorescent images of the same worms were also digitally super-imposed, creating a hybrid image of the C. elegans
(i.e., both fluorescent and transmission). It is evident from these lensfree images that the spatial resolution of our platform is modest compared to a regular lens-based microscope. On the other hand, the main advantages of our platform are its ultra-wide FOV and compact on-chip interface (see ) which might provide an important match for ultra-high throughput screening of C. elegans
samples within automated micro-fluidic systems.
Finally, we would like to also point to an alternative lensfree imaging configuration that can also perform fluorescent imaging of C. elegans
samples on a chip. In this modified configuration (refer to Appendix S1
and Figures S1
for details), we make use of a fiber-optic faceplate inserted underneath the sample substrate to control and tailor the fluorescent PSF of the imaging platform. Compressive decoding of transgenic C. elegans
samples using these altered fluorescent PSFs yields similar imaging results as in , (see Appendix S1
and Figures S3
). This modified configuration can conveniently tailor the fluorescent PSF of the imaging platform to enhance the detection SNR, especially at larger gaps between the object and sensor planes. This could be an important advantage if physical separation between the sample and the sensor-chip is required. Despite this important flexibility, this faceplate based lensfree imaging approach has one limitation: The holographic imaging channel is now significantly distorted since the modes of the fiber-optic faceplate mess the complex spatial frequency content of the holographic field propagating toward the sensor-array. For further details on this modified lensfree on-chip configuration and its C. elegans
imaging results, refer to Appendix S1
In conclusion, we have demonstrated lensfree fluorescent imaging of transgenic C. elegans over an ultra wide field-of-view of >2–8 cm2 with a spatial resolution of ~10 µm. This is the first time that a lensfree on-chip imaging platform has achieved fluorescent imaging of C. elegans. We tested the efficacy of this on-chip imaging approach with different types of opto-electronic sensors to achieve a similar resolution level independent of the imaging chip. Furthermore, we demonstrated that this wide FOV lensfree fluorescent imaging platform can also perform bright-field imaging of the same samples using partially-coherent lensfree digital in-line holography. This unique combination permits ultra-wide field dual-mode imaging of C. elegans which could provide a useful tool for e.g., high-throughput screening applications.