Encrypted login | home

Program Information

Ghost Marker Prediction, Detection and Elimination in Marker-Based Optical Tracking Systems

no image available
G Yan

G Yan*, J Li , K Mittauer , Y Huang , B Lu , C Liu , University of Florida, Gainesville, FL


SU-E-J-165 Sunday 3:00PM - 6:00PM Room: Exhibit Hall

Purpose: In optical tracking system (OTS), ghost marker (GM) limits users' flexibility in marker placement and can lead to incorrect patient setup or system execution failure. This work aims to model the phenomenon of GM and propose robust algorithms for GM prediction, detection and elimination.

Methods: Two GMs occur when two markers and two sensors of the OTS are coplanar. However, when one marker gradually moves out of the plane and breaks the coplanar condition, both GMs briefly remain before they completely disappear. To explain this effect, we expanded line-line intersection definition and the coplanar condition: two lines intersect when their distance is less than threshold Ω and involved points are said to be coplanar within Ω. We fixed one marker and moved the other through the plane with a 0.1 mm step length. Sensor positions were calculated using a back projection method when coplanar condition was strictly satisfied. Ω was determined as line-line distance when ghost marker first appeared. The prediction algorithm checks whether any two given markers are coplanar with the two sensors within Ω; the detection algorithm checks whether any marker is near the intersection of lines connecting sensors and two other markers. A backtracking pattern matching algorithm matches markers to reference markers and eliminates unmatched ones as GMs. The algorithms were validated with two phantoms and 40 SBRT patients using 5~8 markers.

Results: The prediction success rates were 97.7% and 89.6% and the detection success rates were 100% on the two phantoms (1544 & 2045 samples). The matching success rates were 100% on both phantoms and patients.

Conclusion: The proposed model precisely explains the phenomenon of GMs in marker-based OTS. The proposed detection and matching algorithms guarantee complete elimination of GMs, thus avoiding errors in SBRT tracking and providing user maximum flexibility in marker placement.

Contact Email: