Biomedical Engineering Reference
In-Depth Information
lensfree holograms captured with our cell-phone microscope can be significantly
compressed for faster wireless transmission, such that an image corresponding to an
FOV of
5 mm
2
(i.e.,
1 megapixels) can be transmitted using 375 kBytes of data
at
3-4 bits/pixel in a common picture format such as portable network graphics
(PNG) format.
4.5
Lensfree On-Chip Super-Resolution Microscopy
of Biochips
Microscopic biochip imaging devices described in the previous sections have a
lateral resolution of about 1:5m, which proved sufficient for various applications
such as counting of red and white blood cells or waterborne parasites. Note that
this resolution level is still sub-pixel since a pixel size of
2:2misusedto
acquire lensfree holograms in our unit magnification geometry. Nonetheless, a
smaller physical pixel size (e.g., <2m) will allow better sampling of the lensfree
holograms and in turn a finer spatial resolution, since a 2:2-m pixel size may still
lead to undersampling and aliasing of the raw holograms, limiting the resolution of
our microscopy platform.
One approach to overcome this spatial sampling limitation is to cover the
physical pixel with pinholes, which are much smaller than the pixel size. When these
pinholes are arranged in a certain configuration and the object is controllably moved
across the pinholes, finer sampling of the object can be obtained, compared to the
sampling without the pinholes [
23
,
24
]. This scanning-based technique, however,
requires careful fabrication and alignment of the pinholes, and very good control
over the movement of the object, which might be rather limiting for imaging
of heterogeneous samples in, for example, field settings. Refer to Sect.
4.7
for
further discussion on optofluidic imaging of biochips using lensless holographic
approaches.
Using computational on-chip imaging, it is also possible to digitally reach a
smaller effective pixel size with simple mechanical modifications to the devices
described in Sects.
4.3
and
4.4
. Once more, the cheap and widely available
computational power allows transferring some of the burden to the digital domain.
To effectively achieve a smaller pixel size in our on-chip imaging setup, the same
in-line holography configuration is used but now with multiple LEDs, each butt-
coupled to a multimode optical fiber, to sequentially illuminate the objects of interest
(see Fig.
4.9
). The free ends of these fibers are arranged along a line and act as
the illumination pinholes. A simple and inexpensive microcontroller turns on the
LEDs one at a time, which is equivalent to moving the illumination source, causing
a physical shift of the object hologram at the sensor plane. A schematic of this
configuration is shown in Fig.
4.9
. Due to the ratio between the source-object .z
1
/
and object-sensor .z
2
/ distances, the shift of the hologram on the sensor is much
smaller than the distances between the free ends of the fibers.
These spatial shifts between different holograms allow extracting high-resolution
content of the objects through the use of digital pixel super-resolution (PSR)
