Vanderbilt University
Department of
Biomedical Engineering
NCIIA Grant Proposal:
Hadamard Transform Imaging
04 November 2004
Team Members:
Paul Holcomb
Tasha Nalywajko
Melissa Walden
Project Description
Brain tumors are among the most lethal types of cancer, with an average mortality rate of
71% of diagnosed cases.1 Studies have shown a correlation between complete or near-complete
tumor resection and improved prognosis, both in adults and children.2-4 In order for optimal
tumor resection to be performed, however, an imaging modality is needed to distinguish between
normal and cancerous tissue, especially at the tumor margins. Optical spectroscopy, specifically
fluorescence and diffuse reflectance, has been found effective in differentiating between normal
and tumor tissue, both in vitro and in vivo.5-7 Implementation of this technique in spectral
imaging has not been examined for use in guidance of surgical resectioning. Spectral imaging
techniques are currently limited by the low level of autofluorescence of tissue; the weak signal
obtained from current spectral imaging modalities requires long image acquisition and analysis
times in order to resolve the image with the degree of clarity necessary for it to be useful. The
proposed tool, a Hadamard transform-based spectral imaging system using a digital micro-mirror
device (DMD), has the potential to increase signal to noise ratio within the imaging system
substantially, thus reducing the image acquisition and analysis time. This will allow for near
real-time intraoperative imaging of brain tumors and their margins, and will contribute to higher
survival rates for patients.
Currently, three methods are widely used for spectral imaging: line scanning using
motors and a spectrograph, wavelength scanning using filter wheels or electronically tunable
filters, and multiplexing methods using Fourier interferometry. Line-scanning methods, though
ideal for small areas of interest, are relatively slow when large areas must be scanned and the
motors can induce motion artifacts within the images. While wavelength-scanning systems are
rugged and simple to construct, filter wheels only allow spectral measurement at a limited
number of wavelengths within the spectrum. Electronically tunable filters can spectrally resolve
emission over broad wavelength ranges, but they only transmit a narrow band of the emission at
a time, and the transmission within this band generally peaks around 65%. Multiplexing via
Fourier interferometry increases the signal-to-noise ratio over wavelength scanning methods by
collecting 50% of the emission, but the application of the inverse Fourier transform during postprocessing significantly increases the overall acquisition time.
Through application of Hadamard transform multiplexing, many of the problems with the
previous imaging techniques can be resolved. The implementation of this technique is made
possible by the use of a digital micro-mirror device, which consists of an array (1024x768) of
13μm x 13μm micro-mirrors that can be independently positioned to reflect light at two discrete
angles (0º and 12º) relative to the normal axis of the mirror. A Hadamard matrix, which consists
of an array of ones, zeros, and negative ones, can be implemented by positioning these mirrors
for the application of the Hadamard transform to our image. The mirrors are controlled using
CMOS technology, and can change position within 20μs, allowing for rapid repositioning of the
Hadamard matrix during image acquisition. Nearly all of the light passed to the DMD can be
transmitted to the next stage of the system, as the micro-mirrors are almost 100% reflective,
which provides an advantage over previous Hadamard systems which employed liquid-crystal
transmission masks.8 After application of the Hadamard transform, the two reflected images (0º
and 12º, or 1 and -1 of the Hadamard transform) are individually compressed in separate system
“legs” to one dimensional line images using cylindrical optics and then passed to a dispersion
grating. The gratings disperse the spectral components of each line image along an axis
perpendicular to the image, and the resulting two-dimensional (one dimension of emitted
wavelength and one of spatial position within the image) images are collected by a CCD camera.
The number of image samples required for this technique is equal to the order of our Hadamard
matrix (n); after each sample, the Hadamard matrix is shifted to the left one column, with the
first column being added to the end. After all sampling has been accomplished, the inverse
Hadamard transform can be applied and the second spatial dimension of our image recovered.
This inversion consists of addition and subtraction of matrix components, making the process
much faster than the inverse Fourier transform which relies on multiplication and division. By
assigning a color gradient to the image based on the wavelength acquired, the output can be
overlaid on a standard camera image or optical microscope view for real-time use in surgery.
Initially, testing of the Hadamard transform imaging system will be performed using a
single-leg system, examining only one output image from the digital micro-mirror device. A
type of Hadamard matrix called an S-matrix will be used to transform the output data; the Smatrix is essentially a binary form of the Hadamard matrix, allowing for the mirrors to be
represented as either “on” or “off” depending upon their state.9 This initial testing procedure
allows for reduced cost at the outset, as well as a reduction in time required for alignment and
testing. As both outputs of the DMD will eventually be passed through identical optical
compression systems, the design of this single-leg system can be easily extended to the full
Hadamard application. Size of our system is also another important consideration, as the device
must be small enough to be used conveniently in an operating environment. Minimization of the
system will be accomplished through reflective optics, thereby reducing the distance between
optical components.
Many aspects of the Hadamard transform imaging technique make it ideal for use in
spectral imaging. The high transition rate of the DMD micro-mirrors and the rapid application of
the inverse Hadamard transform leads to decreased time between image acquisition and display.
By using the reflective properties of the digital micro-mirror device to implement the Hadamard
matrix in a two leg system, 100% of the light from our image can be used, as opposed to other
spectral imaging techniques which use a maximum of 50%. The use of the Hadamard transform
also leads to increased signal-to-noise ratio (SNR) of the output; theoretically, the SNR of the
image can be improved by a factor of √n, where n is equal to the order of the Hadamard matrix.9
Application of the S-matrix, as is the case for our preliminary testing, still yields a SNR increase
by a factor of √n/2. This signal improvement has been confirmed experimentally in several
studies.8,10-11 A higher signal to noise ratio creates more accurate edges in our image,
guaranteeing better tumor margin visualization.
Market Potential
Because of the near real-time imaging that this design provides, the Hadamard transform
imaging system will be used in conjunction with operating microscopes during brain surgery.
Ideally, this system will augment the operating microscopes already in place, thus reducing cost
to the consumer. As operating microscopes are used by almost all hospitals for microsurgical
procedures, and especially for neurosurgical procedures, the potential market is very large. The
hardware for the Hadamard transform-based spectral imaging system is not a disposable product;
we will be marketing an operating device that should not need replacing, however the
opportunity for technical support is available. Updates to the DMD control and image
processing software allows for possible avenues of profit after product deployment. Testing this
device will require minimal paperwork, as image acquisition is completely non-invasive and has
no foreseeable negative effects on the patient.
References
[1] National Cancer Institute, “Adult Brain Tumor Treatment”,
http://www.nci.nih.gov/cancertopics/pdq/treatment/adultbrain/healthprofessional
[2] Bucci MK et al., “Near complete surgical resection predicts a favorable outcome in pediatric
patients with nonbrainstem, malignant gliomas…” Cancer 101(4):817-24, 2004
[3] Lacroix, Michel et al., “A multivariate analysis of 416 patients with glioblastoma
multiforme: prognosis, extent of resection, and survival”, J Neurosurg 95: 190-198, 2001
[4] Jaing TH et al., “Multivariate analysis of clinical prognostic factors in children with
intracranial ependymomas” J Neurooncol. 68(3):255-61, 2004
[5] Lin WC et al., “Brain tumor demarcation using optical spectroscopy; an in vitro study”
J Biomed Opt. 5(2):214-20, 2000
[6] Lin WC et al., “In vivo brain tumor demarcation using optical spectroscopy”
Photochem Photobiol. 73(4):396-402, 2001
[7] Marcu, Laura et al., “Fluorescence Lifetime Spectroscopy of Glioblastoma Multiforme”
Photochem Photobiol. 80: 98-103, 2004
[8] Hanley QS et al., “Three-dimensional spectral imaging by Hadamard transform spectroscopy
in a programmable array microscope” J Microscopy 197(1): 5-14, 2000
[9] Harwit, Martin, Hadamard Transform Optics, Ch. 1 & 3, New York: Academic Press, 1979.
[10] DeVerse RA et al., “Realization of the Hadamard Multiplex Advantage Using a
Programmable Optical Mask…”, Applied Spectroscopy 54(12): 1751-1758, 2000
[11] Hanley QS et al., “Spectral Imaging in a Programmable Array Microscope by Hadamard
Transform Fluorescence Spectroscopy”, Applied Spectroscopy 53(1): 1-10, 1999
Design Team
Paul Holcomb is a senior biomedical engineering student at Vanderbilt
University. Four years of management experience and previous group
leadership positions have provided Paul with experience in budgeting, time
management, and networking. His previous research experience includes
examination of the role of the EphA2 receptor in tumor angiogenesis with Dr.
Jin Chen at Vanderbilt University Medical Center (2003-present). Besides
assisting with development of the design for the imaging system, Paul will be
organizing the assembly timeline and acting as liaison between the team and
advisory and support staff. After graduating, Paul will be pursuing a PhD in
neural engineering.
Tasha Nalywajko is a senior biomedical engineering and molecular biology
double major at Vanderbilt University. She has performed research for the
past two years in intracellular engineering, focusing on microarrays and
dynamic hybridization. Most recently, her work includes dynamic virus
attachment and detection. Tasha’s skills include image acquisition and
analysis, as well as knowledge of imaging equipment. For this project, Tasha
has contributed ideas and calculations for the optical design, and will assist
with system assembly and data analysis. She hopes to pursue a career in
biotechnology.
Melissa Walden is a senior biomedical engineering student at Vanderbilt
University. She has experience working with computers, including
webpage
design and use of several programming languages. She also has
strong electrical engineering skills, especially bioelectrical components of nerves
and stimuli in the body. Melissa’s role in the project includes design and
design calculations, implementation and assembly of the design, and
maintaining technical contacts for the team web presence. After
graduation, she hopes to attend medical school and pursue rural and family
practice
medicine.
Timeline
September
October
November
December
January
February
February
March
April
May
June
July
August
Design
X
X
X
X
Build (Single-Leg)
Calibration/Testing
Data Analysis
Calibration/Testing
X
X
X
Data Analysis
X
X
X
Miniaturization
Build (Dual-Leg)
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
March and April will also be spent writing and preparing for presentation of our design and
results at the Senior Design Day poster session on April 27, 2005.
Budget
Equipment
Optics
Hardware
Supplies
Travel Expenses
Total
$3,500
($1,200)
($2,300)
$500
$2,000
$6,000
Equipment/Resources Needed
Laboratory facilities have been provided by the Vanderbilt University Biomedical
Engineering department, as have the digital micro-mirror device and CCD camera. The
components required for the single-leg portion of the build have been purchased using grant
money provided by Dr. Anita Mahadevan-Jansen. Additional components for the dual-leg
device and miniaturization have not been purchased.