Evaluation of the impact of resolution-sensitivity tradeoffs on detection performance for SPECT imaging

We evaluated the impact of resolution versus sensitivity tradeoffs on the detection performances of SPECT imaging systems. Images were reconstructed using penalized likelihood with quadratic penalties and the theoretical approximation for the SNR of the Hotelling observer was used as the detection metric. We compared three circularly rotating pinhole SPECT systems with identical detector areas and fields of view (FOVs): The first system consisted of a single pinhole and the projection of the FOV occupied the entire detector area at each view angle. The second system had four pinholes and no overlap between the projections due to increased minification. The third system had nine pinholes with significant overlap between the projections. The systems emphasized resolution, sensitivity and very-high number of counts via overlap, respectively. We simulated 2:1 and 4:1 contrast tumors with 5, 15 and 25 mm diameters within a 100×100×100mm
³
FOV. The four and nine pinhole systems had improved detection performance over the single pinhole system for all simulated contrast and tumor sizes and had almost identical detection performances for the 5 mm tumor. For the larger tumors, the nine pinhole system was slightly better (approx. 10%). These results indicate that the sensitivity advantage gained by using four pinholes instead of one outweighs the loss in resolution and results in improved detection performance. On the other hand, the advantage gained by adding further pinholes with overlapping projections was minimal.