We are developing a Point-Of-Care PET (POC-PET) imaging system platform that consists of one or more movable probe detectors in coincidence with a detector-array behind a patient. The probes are hand movable so that the operator can control the probe trajectory freely to achieve optimal coverage and sensitivity for patient-specific imaging tasks. This platform does not require a conventional full ring geometry, and as such it can be built portable and low cost for bed-side or intraoperative imaging. We developed a prototype that consists of a compact high resolution MPPC detector probe and a half ring of conventional detectors. The probe detector has 20×20 crystals of 0.74×0.74×3.0 mm3 each, in 0.8 mm pitches, read out by a MPPC array. The probe is fixed to a Microscribe device, which tracks the location and orientation of the probe in 3D space as it moves. A fully 3D list-mode TOF (Time-Of-Flight) image reconstruction algorithm has been developed to incorporate the dynamically changing geometry information acquired from the Microscribe. The algorithm is implemented on GPU to achieve fast reconstruction in the order of seconds, under practical count rate situations. Further Monte Carlo simulations show that the resolvability of 4 mm rods under practical contrast ratio could be achieved if the scanning trajectory is well designed.