Motion Correction in Emission Tomography - Correction Techniques in Emission Tomography - page 161

for it. Several methods have been proposed to this end. These include use of
an a priori model of motion, discarding the data from some phases of the
respiratory or cardiac cycles which show the greatest motion, dividing the
emission data into phases and selecting only a particular phase of it, use of
registration techniques to deform the data, etc.
As the respiratory motion is continuous, it is possible to divide the emission
data into several parts, each with minimal motion[17]. This is called gating.
Gating would lead to reducing the amount of motion in each phase; however,
it would also lead to an increase in noise on the reconstructed images due
to less information present in each phase. Therefore, selecting only the best
phase is not the complete solution to the problem at hand. A deformation of
the complete gated data to the target phase is required. This will allow the
full information along with less motion.
A good method for motion correction would be applicable to the PET
data without any prior segmentation or assumptions on the properties of the
different types of tissue present in the body. Before describing the motion
correction techniques, a very short overview of the past attempts at motion
correction in emission tomographic studies is given.
8.2.1 Overview
In those parts of the body where only rigid motion is present, i.e., trans-
lation and rotation, simpler image registration algorithms can be used. One
such set of clinical studies are the brain studies. Picard et al. [52] and Fulton
et al. [28] used video cameras to monitor the skull position externally and
correct for the motion in a post processing step by registration. More recent
attempts also use external monitoring [10] to correct the lines of responses
(LORs) for the motion before reconstruction.
In some studies external motion estimation is used for sorting the events
in accordance with an external motion signal; e.g., Nehmeh et al. [48] used an
external block fastened to the patient's abdomen. The movement of this block
is tracked with video cameras and is used for sorting the data into different
phases. Only the first phase or gate is used for image reconstruction. This type
of method reduces the motion on the images; however, most of the informa-
tion is lost in the unused gates. To get the same statistics as in the ungated
images, a proportionally larger amount of the radiotracer must be used or the
acquisition time has to be prolonged accordingly. Nehmeh et al. also proposed
another method [49] based on the use of an external radioactive point source
which is placed inside the field of view. The emission data is then acquired
in multiple short duration frames. All frames in which the point source falls
within a region of interest (ROI) around the point source are summed up and
used for the reconstruction. This method utilizes more information than the
previous one but still a large part of the events is lost. These methods cannot
be strictly called motion correction; it is rather a selection of frames which
are \good."