Preliminary Investigation of Short-Baseline Stereo Cameras for Motion Detection and Alignment of Pre-existing Attenuation Maps for AdaptiSPECT-C

AdaptiSPECT-C, a multi-pinhole, multi-detector, and stationary brain SPECT imaging system is being developed by the University of Massachusetts Medical School and the University of Arizona. During AdaptiSPECT-C acquisitions, the patient’s head will be positioned entirely inside a hemi-spherical collimator assembly. A popular method to mitigate head motion during the protracted acquisitions is to use external cameras for tracking the motion of the head throughout the acquisition. This motion estimate is then used during SPECT reconstruction to compensate for the motion. Unfortunately, the close proximity of the bore to the patient will occlude the motion tracking camera’s view of the patient from outside the bore, thereby posing a significant challenge to using such methods. Therefore, we propose to use small-baseline stereo depth-sensing cameras placed inside the AdaptiSPECT-C bore, to track patient’s head motion. In this study, using an imaging phantom, we measure the manually generated phantom motion at a distance of about 15 cm from small baseline stereo depth sensing cameras. The motion is measured simultaneously with Optitrack motion tracking system, a commercially available marker based motion tracking system to provide a reference ground truth. Finally, we showed that at close range, we can utilize the 3D surface estimates generated by the small baseline stereo depth sensing cameras, for aligning a preexisting CT to the patient’s SPECT acquisition in order to provide attenuation correction during SPECT reconstruction. Future work will include reconstruction simulations that incorporate motion correction based on actual motion data from the tested cameras as well as providing CT alignment for CT-based attenuation correction.