Siemens develops mathematical approach to lower radiation dose in
computed tomography
25 November 2009
Siemens has developed an innovative method for iterative
reconstruction of CT images.
One of the main challenges of computed tomography (CT) is to provide
excellent image quality while exposing patients to the lowest possible
dose of radiation. Reductions in dose application typically lead to
increased image noise and loss of image quality.
For this reason, Siemens Healthcare has developed what it calls
Iterative Reconstruction in Image Space (IRIS) to generate high-quality
images, acquired with smaller radiation doses. A CT takes a multitude of
X-ray data from different directions and uses the information to
calculate clinical images, which can then be analyzed by physicians.
The newly introduced IRIS algorithm for the reconstruction of
sectional views from CT raw data makes better use of the information
contained in the source data, yet is much faster than previous
approaches to iterative processes in spite of additional reconstruction
steps.
Compared to the current standard method for image reconstruction, the
so-called Filtered Back Projection (FBP), IRIS offers two options to
users of the Siemens procedure: they can either generate the same image
quality as with FBP and reduce the dose by up to 60 percent or they can
maintain the same dose and generate significantly better image quality
than with FBP.
IRIS is currently being tested at several university hospitals. Most
systems of the Somatom Definition product family will be equipped with
the new technology from the second quarter of 2010.
In modern spiral CT devices, patients move through a ring-shaped
tunnel (gantry) at a specific speed, while the X-ray tube
assembly-detector combination continuously rotates around their body.
Mathematical procedures calculate the attenuation coefficient in the
cross-section plane as well as the spatial distribution of density from
the attenuation of the radiation as it passes through the body.
These measuring values, or raw data, are then used to reconstruct
clinical images at different spatial planes, such as axial, frontal,
sagittal etc.
The standard reconstruction method currently in use is Filtered Back
Projection (FBP), an algorithm that converts the raw data into image
data with filtering and back projection to the image plane. This process
involves a compromise between spatial image resolution or image quality,
and image noise. The dose must be increased to lower the image noise and
achieve better image quality.
Iterative reconstruction was first described in the 1970s as a
promising method to generate clinical images with low noise. The image
generation process of this procedure includes a "correction loop", in
which the sectional images are calculated in stages by a gradual
approximation to the actual density distribution. For this purpose, the
system makes an assumption about the density distribution of the tissue
slices to be examined and calculates an output image.
New, synthetic projection data are generated from this output image
and compared to the actual, "real" raw measuring data. If they don't
match, the system will calculate a corresponding correction image to
correct the output image. In a next step, the system will again
synthesize the projection data and compare them to the measured raw
data. This iteration continues until a specified abort criterion is met.
After this process, the corrected image shows improved spatial image
resolution in high-contrast regions, while the image noise in
low-contrast areas is reduced. The image becomes softer in tissue
regions with homogeneous density, while high-contrast tissue boundaries
are maintained. As a consequence, image resolution and image noise are
no longer tied to one another.
One problem associated with the method is the fact that the measuring
system of the CT device must be precisely modeled mathematically during
the computation of the synthetic projection data, which requires immense
computing power.
In addition, a large number of iterations is required. As a
consequence, the calculation time for reconstruction and the computing
capacity requirements increase to such an extent that the procedure
cannot be practically applied in clinical settings.
Until recently, the so-called "statistical iterative reconstruction"
was considered a solution. It avoids the exact mathematical modeling of
the measuring system and drastically reduces the number of iterations to
avoid lengthy computing times.
A large portion of noise is removed on the basis of a simple
statistical correction model, which only focuses on the noise properties
of the measuring data. This aggressive method accelerates the
lower-noise reconstruction of the images considerably, but generates
sectional images that may differ so substantially from the results of
the standard FBP that radiologists are often disturbed by the texture.
In contrast to "statistical iterative reconstruction," the
reconstruction algorithm Iterative Reconstruction in Image Space (IRIS)
by Siemens Healthcare uses a different approach to accelerate the image
reconstruction. The core of the innovative approach is the fact that all
raw data information is transferred from the slow-processing raw data
area to the more efficient image data area in the first reconstruction
cycle.
The resulting "master image" contains finest details, but also
significant image noise, which is removed in the subsequent iterative
steps in the image data area. In this manner, the image is gradually
cleared of image noise and artifacts in small iterative steps that do
not affect the high spatial image resolution.
This eliminates the need for time-consuming back projections. The
novel approach allows Siemens experts to simply construct a highly
precise reflection of the actual properties of the final image from the
raw data of a CT scan with relatively little computing effort. IRIS,
which allows scanning with up to 60% less radiation, can reach the same
signal to noise ratio as Filtered Back Projection (FBP) with a full
dose.
As a consequence, the new algorithm is able to significantly reduce
the radiation dose without quality losses. As an alternative, the
iterative reconstruction method by Siemens can also be used to
substantially increase the image quality of reconstructed images with
the same dose.
This was confirmed by U. Joseph Schoepf, MD, Professor of Radiology
and Cardiology, Director of CT Research and Development at the Medical
University of South Carolina: "Iterative Reconstruction in Image Space
lets me save up to 60 percent of the radiation dose in a number of
routine applications, while maintaining the usual excellent image
quality."
"Radiation protection and dose reduction in CT have been top
priorities of Siemens Healthcare ever since the company came out with
the first computed tomograph (CT) in 1974. We have already introduced a
series of technical innovations to our CT systems that contribute to
dose reduction," explained Dr. Sami Atiya, CEO Computed Tomography at
Siemens Healthcare. "With IRIS we can significantly reduce radiation
exposure in virtually all CT examinations.”