Real-time registration by tracking for MR-guided cardiac interventions

Real-time 2D-3D MR cardiac image registration during respiration using extended Kalman filter predictors

“…Comparison between presented method and that of [2] was inspected during D simulation. Mean of CMM (MCMM) of RT registration process through RTBS series are reported in Table 1 for the selected slices; showing improvement in accuracy, and all are acceptable in clinical condition.…”

“…To evaluate framework performance, Chung et al misalignment metric (CMM) [2] which is simply root mean square difference between the moving image's pixel position in optimum match and assessed match by registration was used in our study. Results of three previously defined simulations summarized below:…”

“…To evaluate similarity of the synthesized RT image and the real ones, noise was assessed for both images and was 0.051 ± 0.02 in real RT images and 0.104 ± 0.031 gray scale values in synthetic RT images. To perform a more realistic simulation compare with the one reported in [2], the respiration cycle period and RT frame rate are assumed 5s and 12FPS, respectively. 450 and 500 samples were used for density estimation in mutual information for RT and prior registration, respectively.…”

“…Drift Simulation (D): 1.66 cycles of respiration with constant drift velocity 0.5mm/s in AP direction and rotational drift 0.1°/s all with a sinusoidal periodic motion as RTBS series [2]. Varying Period Simulation (VP): Five cycles of varying period motion ranging through 5,4,3,4 and 5s, sequentially all with a third order polynomial like periodic motion as RTBS series.…”

“…Chung et al [2] developed a method to detect spatial alignment of RT 2D slices and a prior images using contour tracking during respiration. The authors investigated the cooperation of contour tracking in RT registration.…”