2. Procedure

1
Detecting Patient Motion in SPECT Imaging Using
Stereo Optical Cameras Michael A. Gennert1,2,
Philippe P. Bruyant1, Manoj V.
Narayanan1, Michael A. King1 1University of
Massachusetts Medical School, Worcester, MA
2Worcester Polytechnic Institute, Worcester, MA
Abstract Objectives Patient motion, which causes
artifacts in reconstructed images, can be a
serious problem in SPECT imaging. If patient
motion can be detected and quantified, the
reconstruction algorithm can compensate for the
motion. The goal of this work is to design a
system architecture for detecting, modeling, and
correcting patient motion in SPECT imaging using
information in addition to the emission counts
themselves. Methods Web cameras were mounted
outside a SPECT system to acquire optical images
simultaneously with the emission images. The web
cameras view the patient surface, from which a
surface map may be computed using stereo
techniques. When the patient moves, the surface
map is recomputed. Changes in the surface map
over time enable tracking of the patient surface.
Motion within the patient body is computed by
interpolating from the surface, taking into
account a model of the patient body. When
patient motion is detected, the estimated slices
are reoriented during iterative reconstruction
and notification is made that compensation has
been performed. Results Implementation of the
architecture is currently underway. Conclusions
Further work remains, especially in the area of
using non-isotropic body models for interpolation.
2. Procedure Web cameras were mounted outside a
3-headed IRIX SPECT system to acquire optical
images simultaneously with the gamma emission
images (Figure 1). Before patient data
can be acquired, the optical cameras must be
calibrated to the gamma camera. The calibration
and operation phases are sketched in Figure 2.
Figures 35 show the processing steps during the
operation phase.
1. Introduction Patient motion causes many
problems in SPECT imaging, such as blur and other
motion artifacts. If patient motion is
excessive, it may be necessary to repeat an
acquisition, with consequent cost and
inconvenience. Previous approaches to motion
detection relied on inconsistency checks or
motion tracking, or were limited to rigid body
parts, such as the head. The goal of this work is
to design a more general system architecture for
detecting, modeling, and correcting patient
motion in SPECT imaging using information in
addition to the emission counts themselves.
Left Image
Right Image
3. Project Status Calibration and Image
Acquisition modules are complete. The Stereo
Computation module is currently under
development. Motion Detection, Interpolation,
and Tomographic Reconstruction modules have not
yet been implemented.
Figure 5. Stereo and Motion Geometry. Edges are
matched in left and right images. Changes in
match positions over time indicate patient motion.