Real-Time Modeling of the AdaptiSPECT-C Brain Imaging System for Hardware Evaluation, Acquisition Software Testing, and Adaptation-Rule Development

We have developed and tested a new GPU-based Monte Carlo simulation package for modeling X-ray, gamma-ray, and charged-particle transport. This new package utilizes dynamic CAD models of the imaging system and efficiently encodes all material properties onto the GPU so that all the absorption and scattering cross sections can be computed for each photon trajectory. The simulation software runs entirely on the GPU with no input from the CPU and has achieved over 180 million photons per second on a single laptop computer, which is equivalent to modeling approximately 5 mCi of
^99m
Tc activity in real time. We have used this modeling software to simulate the 24-camera AdaptiSPECT-C brain imaging system, which has a total of 5 pinhole apertures per camera with 4 states per pinhole aperture. Thus, AdaptiSPECT-C can adapt to 1.7 × 10
⁷²
unique configurations. In addition, each camera buffers list-mode measurements that it makes available through a gigabit ethernet channel. The simulation of the AdaptiSPECT-C system models every aspect of the imaging system including source distribution in an XCAT brain phantom, all the independent apertures, interaction in the scintillators, camera statistics, the list-mode processing electronics, and also all 24 TCP list-mode servers on a single computer that faithfully mimics the acquisition hardware. This allows the team to connect the acquisition computer to a separate computer running the simulation software and as far as the acquisition computer is concerned, it is connected to the 24 cameras of the actual imager. We have used this system to evaluate different adaptation rules so that pinhole states can be adjusted during acquisition to maximize the task-performance of the data acquired. We will report on the results of this analysis as well as demonstrate the benefits of real-time SPECT system modeling for system design, analysis, and testing.