Dual gating method for eliminating motion-related inaccuracies in cardiac PET

The aim of this study was to develop gating method that eliminates both respiratory and cardiac motion-related inaccuracies in cardiac PET images utilizing two independent trigger events on PET list mode (LM) data without count loss. The method is based on an idea that two independent triggers are sent periodically into LM data based on respiratory and electrocardiogram (ECG) cycles. This enables separation of dual gated LM data into subgroups which each contain events from certain respiratory and cardiac phase. Two additional assumptions are made: amplitude of respiratory movement as well respiratory triggers are available, and only the last diastolic time interval of cardiac cycle varies due to heart rate instability. An algorithm on MatLab programming environment (Mathworks, INC) for generating dual gated phases from the LM data exported from the DVCT PET/CT scanner General Electrics was developed. The method was tested in phantom studies mimicking realistic respiratory and cardiac motions. Dual gated LM data was grouped to three respiratory phases and four cardiac phases. These dual gated phases were fed back to the database to process 12 dual gated images. Movement of cylindrical (2 mm diameter and 2 mm height) source which rotates (40 mm diameter) and moves axially (20 mm shift) was separated to 12 spatially different segments with well-defined borders on cylindrical surface. Circular images overlapped visually less than 3 mm with each other. Visually examined axially gated images overlapped from 3 to 5 mm. Dual gating method was also tested in one patient study with FDG. In coronal view the surface of myocardium of patient moved 21 mm in cranio-caudal direction when comparing end-expiratory diastolic PET image with end-inspiratory systolic PET image. We have showed that motion correction using the dual gating in cardiac studies improves the accuracy of cardiac PET images. The developed algorithm enables to divide dual gated PET data to several spatially different locations without count loss. Further studies are required to ensure feasibility of the method in human studies.