MIAL

Motion Correction in Medical Imaging

Principal Investigator: Ghassan Hamarneh
Principal Investigator: Greg Mori
PhD Student: Hengameh Mirzaalian
MSc Student: William Ma


This project is in collaboration with Prof. Vesna Sossi at UBC Department of Physics and Astronomy, TRIUMF PET Group [TRIUMF], and MSc. Student Katie Dinelle.

Kinetic Modelling Based Motion Correction

We propose a novel motion correction approach for dynamic emission tomography images that takes advantage of the underlying compartmental models of tracer kinetics.
Our algorithm uses a simultaneous segmentation registration paradigm. The key idea of our approach is that, unlike the standard frame-by-frame (FbF) based registration
methods, we avoid choosing a reference frame and essentially create a reference frame for each time step. The references are generated by assigning time activity curves based
on the estimated kinetic-modeling parameters to the different regions of the segmentation output. We evaluate the goodness of our method compared with the FbF-based registration approach using medical images with known motion corruption parameters. The results indicate superior performance of our method in terms of the accurate estimation of the motion and the kinetic parameters.

 

 Publication

Hengameh Mirzaalian, Ahmed Saad, and Ghassan Hamarneh. Iterative Segmentation and Motion Correction for Dynamic PET Images based on Radioactive Tracer Kinetics. In IEEE workshop on Mathematical Methods for Biomedical Image Analysis (IEEE MMBIA), pages 265-270, 2012 [PDF] [poster]

 

 

Stereo-Vision Based Motion Correction
 

Positron emission tomography (PET), functional Magnetic Resonance Imaging (fMRI), and other functional medical imaging modalities are used to assess brain function in normal and
disease states, but, in general, all are susceptible to head movement. We developed a method for tracking head pose that eliminates the tracker dependence on attaching markers to the head. In particular, a stereo video tracking system, in which left and right high resolution video cameras record head movement and computer vision methods calculate the head’s 3D position, is used.

Constraint

The circular constraint requires features to match across the two base reference images and two input images.

 

Result

Images showing the base points reprojected onto the video frames after applying the rotation and translation estimated by our method. The white dots are the reprojected 3D points for each base feature points. These images show UKF is able to find the correct transformation allowing the base points to hang onto the participant.

 

Video

 

Publication

William Ma, Ghassan Hamarneh, Greg Mori, Katie Dinelle, and Vesna Sossi. Motion Estimation for Functional Medical Imaging Studies using a Stereo Video Head Pose Tracking System. In IEEE Medical Imaging Conference (IEEE MIC), pages 2 pages (accepted), 2008 [PDF]

Site by Faculty of Applied Science web team, Simon Fraser University