Adaptive acquisition protocol design for local CNR maximization in flexible SPECT and PET scanners

We present an adaptive acquisition protocol design technique with the goal of improving local contrast-to-noise ratios (CNR) for flexible SPECT and PET scanners in which the data acquisition parameters at each view can be modulated. Such flexible scanners include rotating SPECT scanners, dual head coincidence imagers, SPECT scanners with variable collimators and those where collimators move in relation to detectors. Variable acquisition parameters include per view scan times (both PET and SPECT) and collimator types and dimensions (SPECT only). In this work, we focus on scan times per view as the parameters of interest. Our technique consists of two steps: (i) acquisition of a scout scan dataset that is reconstructed to determine the region(s)-of-interest (ROI) and the Fisher Information Matrix (FIM) (ii) constrained optimization of a rapidly computable theoretical CNR metric to determine the acquisition protocol that maximizes the approximate CNR at the ROI. The optimization is performed under the constraints that the total scan time is fixed and that the scan time at each view is at least the time spent at that view during scout scan. This ensures that the scout scan dataset becomes part of the final dataset and is not wasted. We describe the theoretical CNR approximation and validate our approach via simulations. We demonstrate significant improvements to local CNR values compared to alternative scan-time allocation approaches.