Combined Correction and Reconstruction Methods - Correction Techniques in Emission Tomography - page 194

format for movement, it is possible to make a temporal rebinning as fine as
desired. If, as in the case of an infrared camera detection, the movement is
known in real time, it is potentially possible to correct each individual detected
LOR for movement [10].

9.3.2 Motion correction during an iterative reconstruction
algorithm
In this section we will describe several approaches to divide the list mode
data in different gates according to the motion information and the incorpo-
ration of motion correction into the reconstruction algorithm.

9.3.2.1 Approaches based on a rigid or ane motion model

A first set of methods is based on data-driven detection and correction of
movement through the adaptation of an iterative reconstruction algorithm.
This methodology has previously been suggested in SPECT for cardiac appli-
cations [34] and for brain imaging [28]. These two techniques decompose in
time all the acquired projections. A first image is reconstructed from one of
these temporal subsets of projections. At each iteration, a subset of projections
is reconstructed, corrected for the movement and the initial image then com-
pensated for the movement is updated by considering these new projections.
At each new iteration, the estimate of the image is registered with respect to
the initial image, allowing the computation of the transformation to be sub-
sequently applied to the projections. Each set of projections corresponding to
the same time is corrected for movement using the computed transformations
and at the end of the reconstruction, a single image is reconstructed free of
any blurring effect due to movement. A similar method was also proposed in
PET using the sinogram format with applications in brain imaging [6].
The second set of methods rely on the use of an external signal to detect
the respiratory motion. Although few techniques use the sinogram format [56],
most correction techniques in pulmonary tumor imaging in PET use the list
mode format. This category of techniques is based on the realignment of all
the LORs detected at the same time interval of the breathing cycle.
In a first step, it is necessary to determine the transformation that will
correct the movement. This transformation can be known through the use an
external detection system, or in the case of a combined PET/CT system, from
the corresponding dynamic CT images synchronized with the respiration [42,
44, 35, 29, 30]. This technique is more accurate than the simple detection of
the respiratory movement by an external sensor such as an infrared camera,
or an impedance belt, in which case the estimate of the true movement of
internal organs from an external signal recorded is not easy. Recently, in PET
cardiac imaging, a technique was proposed to extract the respiratory and
cardiac beating movements directly from the list mode format without the
use of an external device. This method obtains good detection performance as