February 5, 2013
WEST LAFAYETTE, Ind. – Researchers are improving the
performance of technologies ranging from medical CT scanners to digital cameras
using a system of models to extract specific information from huge collections
of data and then reconstructing images like a jigsaw puzzle.
The new approach is called model-based iterative
reconstruction, or MBIR.
“It’s more-or-less how humans solve problems by trial
and error, assessing probability and discarding extraneous information,”
said Charles Bouman, Purdue University’s Michael and Katherine Birck Professor of
Electrical and Computer Engineering and a professor of biomedical engineering.
MBIR has been used in a new CT scanning technology that
exposes patients to one-fourth the radiation of conventional CT scanners. In
consumer electronics, a new camera has been introduced that allows the user to
focus the picture after it has been taken.
“These innovations are the result of 20 years of
research globally to develop iterative reconstruction,” Bouman said.
“We are just scratching the surface. As the research community builds more
accurate models, we can extract more information to get better results.”
In medical CT scanners, the reduction of radiation
exposure is due to increased efficiency achieved via the models and algorithms.
MBIR reduces “noise” in the data, providing greater clarity that allows
the radiologist or radiological technician to scan the patient at a lower
dosage, Bouman said.
“It’s like having night-vision goggles,” he
said. “They enable you to see in very low light, just as MBIR allows you
to take high-quality CT scans with a low-power X-ray source.”
Researchers also have used the approach to improve the
quality of images taken with an electron microscope. New findings are detailed
in a research paper being presented during the Electronic Imaging 2013 conference
in San Francisco this week.
Traditionally, imaging sensors and software are designed
to detect and measure a particular property. The new approach does the inverse,
collecting huge quantities of data and later culling specific information from
this pool of information using specialized models and algorithms.
“We abandon the idea of purity – collecting precisely
what we need,” Bouman said. “Instead, let’s take all the measurements
we possibly can and then later extract what we want. This increases the
envelope of what you can do enormously.”
Purdue, the University of Notre Dame and GE Healthcare
used MBIR to create Veo, a new CT scanning technology that enables physicians
to diagnose patients with high-clarity images at previously unattainable low
radiation dose levels. The technology has been shown to reduce radiation
exposure by 78 percent.
“If you can get diagnostically usable scans at such
low dosages this opens up the potential to do large-scale screening for things
like lung cancer,” Bouman said. “You open up entirely new clinical
applications because the dosage is so low.”
A CT scanner is far better at diagnosing disease than planar
X-rays because it provides a three-dimensional picture of the tissue. However,
conventional CT scanners emit too much radiation to merit wider diagnostic use.
“But as the dosage goes down, the risk-benefit
tradeoff for screening will become much more favorable,” Bouman said.
“For electron microscopy, the principle advantage is higher resolutions,
but there is also some advantage in reduction of electron dosage, which can
damage the sample.”
The research to develop Veo has been a team effort with
Ken Sauer, an associate professor of electrical engineering at Notre Dame, in
collaboration with Jean-Baptiste Thibault, Jiang Hsieh and Zhou Yu. Thibault
and Yu worked on the technology as graduate assistants under Bouman and Sauer
and both currently work for GE Healthcare.
“And, there are lots of other people doing similar
and related research at other universities and research labs around the
world,” Bouman said. “Ultimately,
3-D X-ray CT images might require little more dosage than old-fashion planar
chest X-rays. This would allow CT to be used for medical screening without
significant adverse effects.”
In the electron microscope research, MBIR was used to take
images of tiny beads called aluminum nanoparticles.
“We are getting reconstruction quality that’s
dramatically better than was possible before, and we think we can improve it
even further,” Bouman said.
Improved resolution could help researchers design the next
generation of nanocomposites for applications such as fuel cells and transparent
The paper was authored by Purdue doctoral student
Singanallur Venkatakrishnan; U.S. Air Force Research Laboratory researchers
Lawrence Drummy and Jeff Simmons; Michael Jackson, a researcher from BlueQuartz
Software; Carnegie Mellon University researcher Marc De Graef; and Bouman. A
tutorial article (pdf)
also appeared in January in the journal Current
The models and algorithms in MBIR apply probability
computations to extract the correct information, much as people use logical
assumptions to draw conclusions.
“You search all possible data to find what you are
looking for,” Bouman said. “This is how people solve problems. You
saw Bob yesterday at the store; you wonder where he was coming from. Well, you
determine that he was probably coming from work because you have some
probabilistic models in your mind. You know he probably wasn’t coming from San
Francisco because Bob doesn’t go to San Francisco very often, etc.”
MBIR also could bring more affordable CT scanners for
airport screening. In conventional scanners, an X-ray source rotates at high
speed around a chamber, capturing cross section images of luggage placed inside
the chamber. However, MBIR could enable the machines to be simplified by
eliminating the need for the rotating mechanism.
Future research includes work to use iterative
reconstruction to study materials. Purdue is part of a new Multi-University
Research Initiative funded by the U.S. Air Force and led by De Graef.
Researchers will use the method to study the structure of materials, work that
could lead to development of next-generation materials.
Writer: Emil Venere, 765-494-4709, [email protected]
Bouman, 765-494-0340, [email protected]
Note to Journalists: An electronic copy of the
SPIE research paper is available from Emil Venere, Purdue News Service, at
765-494-4709, [email protected]
Model Based Iterative
Reconstruction for Bright Field Electron Tomography
V. Venkatakrishnan a, Lawrence F. Drummy b, Marc De Graef
c, Jeﬀ P. Simmons b,
and Charles A. Bouman a
Purdue University; b Air Force Research Lab; c Carnegie Mellon University
Field (BF) electron tomography (ET) has been widely used in the life sciences
to characterize biological specimens in 3-D. While BF-ET is the dominant
modality in the life sciences, it has been generally avoided in the physical
sciences due to anomalous measurements in the data due to a phenomenon called
“Bragg scatter” – visible when crystalline samples are imaged. These
measurements cause undesirable artifacts in the reconstruction when the typical
algorithms such as Filtered Back Projection (FBP) and Simultaneous Iterative
Reconstruction Technique (SIRT) are applied to the data. Model based iterative
reconstruction (MBIR) provides a powerful framework for tomographic
reconstruction that incorporates a model for data acquisition, noise in the
measurement and a model for the object to obtain reconstructions that are
qualitatively superior and quantitatively accurate. In this paper we present a
novel MBIR algorithm for BF-ET which accounts for the presence of anomalous
measurements from Bragg scatter in the data during the iterative
reconstruction. Our method accounts for the anomalies by formulating the
reconstruction as minimizing a cost function which rejects measurements that
deviate significantly from the typical Beer’s law model widely assumed for
BF-ET. Results on simulated as well as real data show that our method can
dramatically improve the reconstructions compared to FBP and MBIR without
anomaly rejection, suppressing the artifacts due to the Bragg anomalies.