Comparison of Algorithms for Prediction of Respiratory Motion

Tejas Nair¹, H. Michael Gach¹

MRI techniques that have long acquisition deadtimes and are highly sensitive to motion (e.g., subtraction techniques like arterial spin labeling) can greatly benefit from predictive motion correction. Two Kalman filter based prediction models [the constant velocity (CV) and the interacting multiple model (IMM)] and a weighted fourier linear combiner (WFLC) algorithm were evaluated for respiratory motion prediction. The CV model, IMM and WFLC predicted organ motion 1 s into the future with a root mean square error of >6.6 mm, >5.3 mm and >1.5 mm, respectively. The CV model is computationally the fastest algorithm followed by IMM and WFLC.