Transcript 48x36 Poster Template - Bourns College of Engineering

Image Processing in Spectral Domain Optical Coherence Tomography (SD-OCT)
Vasilios Morikis1,2, Dan DeLahunta1,3, Md. Shahidul Islam4, Christian M. Oh4, Hyle Park4
1 Bioengineering Research Institute for Technical Excellence, UC Riverside
2 Department of Nanotechnology, UC San Diego
3 Department of Physics, University of Rochester
4 Department of Bioengineering, UC Riverside
Abstract
Optical Coherence Tomography (OCT) is an optical imaging
technique based on low coherence interferometry of light
waves. This method is mostly useful for obtaining high
resolution cross-sectional images of biological tissues at a
high speed. OCT is advantageous for some of its features
which include non-invasive procedures, minimal contact with
tissues, use of non toxic dyes, good lateral and axial
resolution of images and better in-depth imaging than other
optical methods. Because of these features, OCT has been
an important imaging method in Ophthalmology,
Dermatology, Cardiovascular imaging, Neuroimaging and
many other fields. An OCT system utilizes low coherent light
source and an optical set up to produce interference patterns
in the spectrometer and these interference patterns, termed
as sample depth profiles, are later processed in the
computer to obtain the final image of the sample. This project
is looking at the post processing steps in an OCT system and
the goal is to analyze the data obtained from the
spectrometer and perform the image processing techniques
to generate the final image. A Fourier transform of the raw
data from the spectrometer has a high degree of artifact. So,
in order to remove the noises and obtained high quality
images, we need to extensive post-processing of the data.
Some of the basic post processing steps includes reading
the image as a matrix, flipping the matrix if necessary, zero
padding, interpolation, and fast Fourier transform (FFT).
Once the images are processed, they can be arranged
altogether to generate a cross sectional image of the sample.
Optical Coherence Tomography
Project Overview and Methodoly
Conclusions
This project is a mathematical focusing of raw data obtained from an OCT system. A
series of steps must be done prior to an FFT to increase the signal to noise ratio (SNR)
and produce a clear high resolution image.
Read the
Image
Flip Matrix
(if necessary)
The processes are the ones illustrated in Figure 3. After MATLAB reads the image, it
extracts a matrix from the image. Depending on which camera the image is from, it may
have to be flipped.
Focusing
The
next
step
is
to
expand
the
matrix
by
Lens
Fig. 3
adding zeroes so that a more accurate
Diffraction
Setup of interpolation may occur. Interpolation is used
Grating
the OCT to find linearly spaced values so that an FFT
system
Fast Line
can
be
performed,
in
this
case
it
takes
K
Scan
Cameras
(wave number) and makes it linear.
Collimator
Interpolate
Zero
Padding
Polarized
beamsplitter
cube
Fig. 11 Images obtained from the
1310 nanometer system of a thin
slice of apple (image width: 100
microns, image height: 500 microns
approximately). The one on the left is
completely unprocessed while the
one on the right is mathematically
focused using the parameters from
this project.
Display
Image
FFT
Fig. 4 Process in which the MATLAB code
analyzes the data produced from OCT.
Fast Fourier transforms are used to switch one complex variable to another one,
in this case it transforms K into actual space.
An objective of this project is to find better depth profiles by adjusting
parameters in the MATLAB code. More specifically the incident angle, the focal
length, and the wavelength of the system.
Fig. 5 Equation that
incorporates the
parameters: incident
angle, grating spacing,
focal length, and initial
wavelength.
Results
Fig. 6 Raw data obtained from the straight camera when the reference
and sample arm are 600 microns apart and
Now that the parameters are working images can be
retrieved from the camera and focused mathematically to
obtain much clearer images.
The raw data taken of just a mirror, with no sample, from the
camera comes in (Figure 6) and is then read by the
MATLAB program, when no processing steps are done we
get a graph that looks like Figure 7. To create an accurate
image the point spread function should be narrow and high
(ignore the noise in the middle). As you can see Figure 7 is
wide and that is why at the top of the image we see a thick
black blur. Whereas the image of a mirror should me a
single thing line.
Once all the processing steps are complete (Figure 8) we
get a much sharper point spread function which produces
the desired result of a thin black line. As distance between
reference and sample arm decreases the point spread
function shifts right.
In Figure 11 the unprocessed image is quite blurry but after
mathematically focusing the image, the blurs become very
sharp objects. The majority of an apple is water while the
actual structure of the apple is a loose network so the dark
spots would be the actual flesh of the apple while all the
small white areas woven between the dark spots would be
the juice of the apple. All the grey area beneath the apple
would be air.
Future Work
Now that the systems are
producing images,
it is
possible to venture into many
fields. Such as looking at
damaged rat nerves and, even
further down the line, human
tissue and nerves.
Fig. 12 Diagram of the
sciatic nerve of a rat, and a
potential crush site.
Fig. 1 Outline of OCT system
In Optical Coherence Tomography tissue is placed under a
light source. The light is then reflected back, an
interferometer must be used to detect the extremely short
time delays.
A two dimensional cross section or
a three dimensional volume can be
formed by scanning the beam
across the tissue.
Fig. 7 Unprocessed Distance vs. Intensity graph and the corresponding
image of Figure 6.
Fig. 9 Processed Distance vs. Intensity graph and the corresponding
image when reference arm and sample arm are 400 microns apart.
Fig. 2 Sample image
The 1310 nanometer system we are utilizes a polarizing
beam splitter cube and two cameras to acquire data. That
way we can split the detected light and reconstruct the
polarization state of light returning from the sample.
POSTER TEMPLATE BY:
www.PosterPresentations.com
Fig. 13 OCT image of
a crushed sciatic rat
nerve one day after
crush was applied,
white areas are
inflammation
Fig. 8 Processed Distance vs. Intensity graph and the corresponding
image of Figure 6.
Fig. 10
Incident Angle
Parameters
used to create Grating Spacing
better point
Focal Length
spread
Wavelength
functions
PixelWidth
Side Camera
49*pi/180
1.0e-3/1145
9.20E-02
1.35E-06
2.50E-05
Straight Camera
51*pi/180;
1.0e-3/1145
9.50E-02
1.35E-06
2.50E-05
Acknowledgements
The authors thank NSF and the UC Riverside BRITE program for
funding, as well as the University of California, Riverside and
NIH (R00 EB007241), as well as the entire BIONIL group for their
guidance.