Download Speckle Noise Reduction in Optical Coherence Tomography By

Speckle Noise Reduction in Optical
Coherence Tomography
By:
Marisse Foronda
Rishi Matani
Hardik Mehta
Arthur Ortega
Bioengineering 175 – Senior Design
University of California Riverside
January 9, 2010
Executive Summary
The presence of coherent speckle in optical coherence tomography (OCT) images can
obscure identification of small or thin tissue structures. We present a method of reducing the
effect of speckle in a multifunctional spectral domain OCT system by modifying the wavefront
in between consecutive depth scans and consequently the speckle pattern using a microdeformable mirror placed in the sample arm beam path. We hope that by adjusting the wavefront
modification between each depth profile acquisition and subsequently incoherently averaging
small numbers of adjacent depth profiles, we can achieve a considerable reduction in speckle
contrast without a significant increase in the phase noise floor or overall acquisition time. We
will demonstrate the results of our technique on biological samples. As coherence imaging
continues to emerge as the prevalent player in the field of non invasive diagnosis imaging, the
importance of resolution cannot be understated. The deformable mirror will revolutionize
imaging results and eventually become a standard component in various imaging technologies.
Introduction
Optical Coherence Tomography (OCT) is an emerging field in biomedical imaging. OCT
is minimally or non-invasive, depending on the application, and renders high resolution crosssectional images that can be used to visualize sub-surface microstructures in biological tissues;
yet, with these high resolution images, speckle formation triggers a problem. The grainy and
high contrast nature of coherent speckle in OCT images makes detection of boundaries as well as
small or thin structures difficult. A fully developed speckle pattern forms due to coherent
multiple backscattering and forward scattering occurs from a single focal volume in biological
tissue. Hence, the main purpose of this project is to build a system that will reduce the amount of
speckle formation within a single image. Currently, there have been two successful methods that
show desired results in reduction of speckle formation. One method used was digital filters
during post processing of data. The other method used was physical techniques in which multiple
images were taken from the same location and incoherently averaged to reduce the appearance of
speckle noise. However, although these methods produced promising results, they failed to
acquire real time analysis. Real time analysis is essential for time efficiency and overall progress.
Our system will achieve significant speckle reduction with no increase in acquisition time and
utilize real time analysis in data acquisition while reducing the amount of speckle formation in
the images gathered. We have searched for commercial products but were unable to find any
available. We were able to locate a pending patent application (Park BH, de Boer JF, “System,
arrangement and process for providing speckle reductions using wave front modulation for
optical coherence tomography,” pending US patent application 60/760,592, filed 1/20/2006.).
2.Project Description
2.1 Objective
The objective of this project is to present a method of reducing the effect of speckle in
multifunctional spectral domain OCT by modifying the wavefront and consequently the speckle
pattern using a micro-deformable mirror placed in the sample arm beam path. The optical
component (deformable mirror) inserted in the sample arm beam path will modify the wavefront
of light incident on and reflecting from the tissue. With the ability to manually control the
conformation of the deformable mirror, the resulting observed speckle pattern can be altered,
thereby increasing the quality of sample images. The resulting changes in observed speckle with
be quantified using the speckle contrast ratio, defined as the standard deviation intensity divided
by the mean intensity over a given region.
2.2 Methods
Figure 1: Detailed scheme of OCT system. SLD: superluminescent diode source, P:polarizer,
PM: polarization modulator, C: circulator, BS: beam splitter, PC: polarization controller, R:
reference arm mirror, μDM: micro-deformable mirror, G:galvonometric x-y scanner, S:sample,
G: transmission grating, PBS: polarizing beamsplitter cube, LCS: line scan camera.
The implementation of the speckle reducing technique includes a multifunctional
spectral-domain OCT system. We will be using chicken muscle tissue as our sample.The
broadband light source will be capable of acquiring intensity, birefringence, and flow
information at fixed rates. The micro-deformable mirror is inserted in the sample arm beam path
which is comprised of 12 x 12 controllable actuators with maximum vertical stroke of 3.4
A
gold membrane mirror is attached to each actuator, allowing control of the entire mirror
configuration. Electronic alteration of the deformable mirror’s configuration to a multitude of
different patterns, while acquiring successive depth profiles and averaging these profiles, will
contribute to speckle reduction.
Modification of the wavefront, both incident on and returning from the tissue sample
in such a way that adjacent points will constructively and destructively interfere differently and
create unique speckle patterns. Eventually, once speckle decorrelation is achieved, speckle
reduction must be quantified to compare results accurately. Speckle will be quantified using the
speckle contrast ratio, defined as the standard deviation intensity divided by the mean intensity
over a given region. We will calculate this value over small regions in order to prevent changes
in the tissue microstructure affecting our results.
The mirror is a smooth membrane spanning all the actuators, and so will not in practice
contain abrupt changes, but instead exhibit smoother transitions between peaks and troughs. As a
result, there will be loss in signal-to-noise ratio (SNR) due to a lowered coupling efficiency back
into the fiber. An increased distance between peaks and troughs on the mirror, due to a greater
difference in applied voltages to adjacent actuators, will result in both higher speckle
decorrelation and higher scattering loss. This means that increased speckle reduction using the
present technique will invariably be associated with a greater decrease in SNR. Note that using a
flat mirror with no deformation applied will result in the highest SNR attainable with the system.
The tradeoff between decorrelation and SNR can be quantified by varying the deformation
magnitude, or the voltage difference applied to the mirror between the peaks and troughs of the
pattern, while taking images of the same tissue area. We calculate for each deformation
magnitude the signal-to-noise ratio and the normalized correlation coefficient with an image
taken with a flat mirror. The latter, representing the degree of speckle decorrelation was found by
taking a representative area of the patterned mirror and flat mirror images, and using the equation
Equation 1:
Correlatio n Coefficien t 
max corr Flat , Patterned
max corr Flat , Flat  max corr Patterned, Patterned
where corr( A, B)  FFT FFTA FFTB is the standard cross correlation function, all Fourier
transforms are two dimensional, and  is the dot product operator. The function produces a value
between 0 and 1, with 1 denoting maximum correlation and 0 no correlation at all.
1
We will quantify the reduction in speckle using the speckle contrast ratio, defined in
general terms as the standard deviation intensity divided by the mean intensity over a given
region. The calculated result can often be affected by changes in tissue microstructure, especially
if done over larger areas. To deal with this issue, we plan to partition each image into smaller
regions (to be determined) and find the standard deviation and mean intensity in each region, and
subsequently averaged the calculated speckle contrast across all regions in the image. This will
provide a more accurate quantification of variations in intensity caused by speckle and not by
tissue variations.
The scattering properties of human tissue vary greatly based on tissue type and location,
and as a result will present a wide range of scatterer densities. It must be demonstrated that the
speckle reduction technique does not lose effectiveness when faced with density variations in
order to make sure the technique works universally. For this reason, several tissue phantoms will
be prepared by thoroughly mixing heated agar gel, 5-micron microspheres, and various
concentrations of intravenous lipid, then adding the mixtures to 1 cm-diameter wells for OCT
imaging. The agar gel maintained the solidity of the medium, preventing speckle decorrelation
from particle flow, the microspheres provided larger scatterers for visualization and focusing
purposes, and the intravenous lipid provided the bulk of the scatterers and acted as the tissue
model. Phantoms containing varying intravenous lipid concentrations will be imaged with mirror
deformations spanning over increased voltages, and the reduction in speckle contrast will be
calculated.
Using the correlation coefficient mentioned above, we can quantify speckle contrast
reduction as a percentage compared to the standard image. Prior work has achieved low
percentage of speckle reduction, so we hope to achieve over 50% reduction, but this number will
vary over different tissue samples. We will know that progress is being made as the speckle
reduction correlation increases towards our theoretical maximum that can be calculated based on
the mirror’s pattern (equation 1).
3.Budget
ITEMS
BOUGHT
Deformabale
Mirror
(Software)
{DM140-35UM01}
12*18 Bread Board
{MB1218}
Rotational Mount
{XYR1/M}
Pedestal Post
{RS1.5P8E}
Pillar Post
Extensions
{RS2}
Adaptive Post
Galvo
{GVS002}
Forks
{9947}
Mirror Mount
{9807}
Collimator
{PAF-X-7-B}
Fiber Optic Patch
Cable
{P3-SMF28-FC-2}
Agar {400402500}
Titanium Oxide
{277370010}
Biological Samples
NEED TO
BUY
Qty
COST
X
1
17500
X
1
208.7
X
1
550
X
5
113.75
X
5
100
X
X
5
1
1895
X
10
210
X
5
490
X
1
428.4
X
1
59.9
1 (250g)
1 (kg)
64.52
40.44
X
X
X
21660.71
Conclusion
Several existing methods for speckle reduction have been implemented in various types
of OCT. The simplest method is to average a large number of acquired OCT images of the same
sample. This requires a tremendous amount of time, work and storage since a multitude of
images exist, most of which have highly correlated speckle patterns. Other methods, such as
polarization diversity, spatial compounding, and frequency compounding all offer unique
methods of speckle reduction but are often limited in their capabilities. Most commonly, postprocessing methods have been used to reduce speckle, but these techniques require extensive
computation and are specific to certain types of OCT.
Below are the key advantages of our technique that make our project unique.

Speckle reduction can be achieved within a single image by averaging small numbers of
depth scans with different speckle patterns, reducing the amount of data to be acquired.

This technology can be easily implemented into pre-existing OCT systems (as well as
other forms of OCT) by attachment of deformable mirror in sample arm beam path

Eliminates the need for complex post processing computation
The market for our product is limitless as the deformable mirror can essentially be implemented
into any type of coherence imaging technology. Many types of imaging, including OCT, are
emerging technologies that are often limited by their resolution. The deformable mirror is
especially unique in its capability to easily incorporate into the sample arm beam path of existing
imaging systems, opening the product to a wide array of possibilities. The obvious issue is the
steep cost of the deformable mirror which can range up to $20,000. Understanding that the bulk
of the cost stems from the mirrors ability to be customized down to the pixel, our study hopes to
show the consistent improvement in image resolution (speckle reduction) from certain patterns of
the deformable mirror. This will allow for the production of a prototype mirror that consists of a
few predetermined mirror configurations, resulting in a much more inexpensive product.