Factors affecting repeatability of activity quantification in SPECT/CT

Objectives For many clinical applications of SPECT/CT, including treatment planning and monitoring therapeutic response, it is important to understand the repeatability of activity estimates. The aim of this work was to evaluate instrumentation, reconstruction, and analysis components of repeatability for state-of-the-art SPECT/CT systems.

Methods We used both in-air and in-water measurements of Ba-133 sources to assess the repeatability. This radionuclide was chosen because of its long half-life and its similar emission spectrum compared to I-131. The sensitivity of each system was estimated with a static acquisition of a 4 mm diameter sealed disk source. Sensitivity was measured 10 times without movement of the source and 4 times at 2 week intervals. The phantom consisted of 4 rods filled with Ba-133, with diameter ranging from 0.8 cm to 2.8 cm placed in a water-filled cylinder. SPECT/CT data were acquired 5 times consecutively and 4 times at 2 week intervals. Images were reconstructed with compensation for attenuation, scatter and collimator-detector response. Total activity was measured using a VOI defined by thresholding the CT image to give the true rod volume (VOI1) and (VOI2) expanded by 1 voxel in all directions. Repeatability of sensitivity and total activity in the rods was assessed by the coefficient of variation (COV).

Results Sensitivity from consecutive measurements had COVs of 0.73% and 0.79% for the two cameras. The COVs over 2 week intervals were 0.85% and 0.90%, respectively. The ranges of COVs for the various rods were 1.3% to 5.1% and 1.4-8.2% for the measurements made consecutively and at 2 week intervals, respectively. Corresponding results for VOI2 were 1.4-2.2% and from 1.6-2.7%.

Conclusions Instrumentation repeatability was good over both short and 2 months intervals. This suggests that sensitivity measurements do not need to be made for each patient. The poorest repeatability was for small objects and was due to variation in voxelization of the objects, indicating the need to use appropriate voxels and carefully define VOIs.