Evaluation of a cardiac SPECT reconstruction approach for patient dose reduction

1758 Objectives We’ve developed a three-dimensional (3D) reconstruction technique (nSPEED) that enables acquisition time reduction, or alternatively patient dose reduction, by a factor of two through improved image quality over the conventional FBP or 2D reconstruction. In a revised approach (nSPEED 2), we incorporated a 3D Bayesian technique with an edge-preserving prior and a weighted-Gaussian filter to further improve image quality and reduce patient dose. Methods We applied both nSPEED and nSPEED 2 to anthropomorphic phantom studies with known defects for defect contrast evaluation and 88 patient studies that went through the standard same day rest/stress scans for image quality evaluation. An experienced nuclear medicine technologist graded the patient images in terms of resolution, smoothness, myocardium-to-chamber contrast, gut activity interference with the myocardium, and the overall image quality. During the grading, the technologist was blinded of the processing approach. Results With nSPEED 2 defect contrast was equivalent to or better than with nSPEED: 19.4% vs 17.2% (an anterior defect) and 17.4% vs 17.0% (an inferior defect) in phantom 1, 49.7% vs 45.1% (an anterior defect) and 32.9% vs 28.8% (a lateral defect) in phantom 2, and 32.3% vs 29.3% (a lateral defect) in phantom 3. The grading results showed significant improvement in all but one image aspect using nSPEED 2 over nSPEED (see table). Moreover, the overall image quality of the rest studies using nSPEED 2 was better than the stress studies using nSPEED (3.52 vs 3.25, p Conclusions The revised nSPEED reconstruction preserved or improved defect contrast and significantly improved image quality. Patient studies suggested potential dose deduction by a factor of six when using the revised nSPEED for cardiac perfusion SPECT