Maximum-likelihood reconstruction with ordered subsets in bone SPECT

This study was aimed at determining whether the ordered-subset expectation maximum (OSEM) is more effective than filtered backprojection (FBP) for bone SPECT in the routine clinical context.

Fifty-seven consecutive bone SPECT studies were analyzed. They included pelvic and lumbar spine, thoracolumbar spine, head and neck, feet and shoulders. A 64-projection SPECT study was acquired over 360 degrees by single-head cameras 2-3 h after the injection of 750 MBq 99mTc-methylene diphosphonate. Three observers compared the OSEM and FBP reconstructed images.

Streak artifacts, always present with FBP, were rarely generated with the OSEM. When present (n = 24), artifacts associated with negative values near hyperactivities in FBP were not generated with the OSEM in 67% of the cases (n = 16), permitting a satisfactory interpretation of these regions. In half of the other cases (17%, n = 4/24), interpretation was precluded. In only one case did the three observers agree that more hyperactivities were seen with the OSEM. Ninety-six percent of the OSEM pictures were superior or equal to FBP for anatomic resolution and were clearly better in 12% of the cases. The extent of the lesion with the OSEM seemed better or equally defined in 96% and clearly better in 14% of the cases. The low-activity regions were better or equally visualized in all cases and were clearly better seen in 23% of the cases. The quality of the pictures was found to be better or superior with the OSEM in 98% of the cases and definitely better in 65% of the cases.

Replacement of FBP by the OSEM in bone SPECT would be beneficial to clinical practice.