Transcript Image processing in Spectral Domain Optical Coherence

Vasilios Aris Morikis
Dan DeLahunta
Dr. Hyle Park, Ph.D.

Optical Coherence Tomography
◦ An Overview of OCT

System Setup
◦ Sample Arm
◦ Galvanometer

Project Overview
◦ Methodology
◦ Results
◦ Conclusions


High resolution sub-surface imaging
Non-invasive
◦ Not harmful to subject

Potential in many fields
◦ Ophthalmology (RNFL thickness, AMD)
◦ Dermatology (photoaging, BCC detection)
◦ Cardiology (assessment of vulnerable
plaques)
◦ Gastroenterology (Barrett’s esophagus)

Time delay between reflected light is measured to determine depth
of the reflecting structure
◦ Due to the short time delays between signals OCT must use an
interferometer to detect the reflected light.


Interference fringes are formed when the sample and reference arms
are within a small range.
A depth profile is formed by the detection of the interference pattern
between the reference and sample arm as the reference arm is
scanned.


The intensity of the depth
profile is encoded on a
logarithmic scale.
A 2D cross section or even a
3D volume can made by
scanning the beam across
the sample.

Helped to
construct the
Sample Arm.

Built the box to
power and
control the Galvo

Video of the
Galvo moving


Develop analysis/processing code in
MATLAB
Objective: Mathematically focus raw data
obtained from the 1310 nanometer
system.
◦ Adjust the incident angle, focal length, and
the wavelength.
◦ Increase the signal to noise ratio (SNR) to
produce high resolution image.
Focusing
Lens
Polarized
beamsplitter
cube
Diffraction
Grating
Collimator
Fast Line
Scan
Cameras
Read the
Image
Flip Matrix
(if necessary)
Interpolate
Zero Padding
FFT
Display Image

Raw data obtained
when the reference
and sample arm are
600 microns apart.
Image taken of the
mirror.
Intensity

Pixel Number



Creates a blurred black line
when the actual image is
formed.
Completely unprocessed
data.
To create accurate image
point spread function should
be narrow and high (ignore
all the noise in the middle).


Splits the matrix and adds many 0’s in
Fourier space.
Doubles the size of the original graph.
Intensity
Intensity
◦ Used to increase the point density to interpolate
more accurately.
Pixel Number
Pixel Number


Used to find remap the data linearly in
wave number (k) to improve the results
of a subsequent FFT
Takes the Intensity vs. Pixel number
graph and Intensity vs. k.

Fourier transform switches one complex
valued function into another.
◦ Transforming k (wave number) into actual space.
Side Camera
Straight Camera
Incident Angle
49 degrees
51 degrees
Grating Spacing
1.0e-3/1145
1.0e-3/1145
Focal Length
92 millimeters
95 millimeters
Wavelength
1350 nanometers
1350 nanometers
Pixel Width
25 micrometers
25 micrometers

Now that the parameters are correct a
much more focused image is created.
◦ Dark line at the top is thin and not blurry

Image obtained form the
1310 nanometer system
before processing (left)
and after processing
(right).
◦ Image width:100 microns
◦ Image height: 500 microns

Very first image acquired
with either system.


I would like to thank NSF and the UC
Riverside BRITE program for funding, as
well as the University of California,
Riverside and NIH (R00 EB007241), and Dr.
Hyle Park and the rest of the Park Research
Group for their guidance.
Mr. Jun Wang for organizing and Dr. Victor
Rodgers for directing the program.