Quantitative accuracy of tumor activity estimates from quantitative In-111 SPECT images

Objectives Accurate pre-therapy prediction of tumor response in targeted radionuclide therapy (TRT) would be useful for optimizing therapy. This requires accurate dosimetry, which in turn relies on accurate tumor activity estimates. In this work, we investigated the effects of various factors including tumor size, position, and tumor-to-background ratio on accuracy and precision of tumor activity estimates from In-111 SPECT images.

Methods We generated a set of 3 tumor sizes in each of 3 tumor shapes. Tumors were placed at various positions in the NCAT phantom including inside or on the boundaries of various organs. Projection data of the organs and tumors were computed using Monte Carlo simulation. The projections were scaled and summed to model various activity distributions and 2 tumor-to-background ratios (TBRs), 4:1 and 8:1. Multiple noise realizations of the projection data were reconstructed using OS-EM with compensation for attenuation, scatter and full collimator-detector response. The accuracy and precision of tumor activities estimated from reconstructed images were calculated.

Results For spherical tumors, the mean percent errors of activity estimates for tumors with diameters equal to or larger than the system resolution (~=1.8cm) and both TBRs were less than 6% with standard deviations of less than 2% after 50 iterations. For smaller tumors (diameter = 0.9 cm), the mean percent errors were larger than 70% and still improving after 50 iterations, but standard deviations were more than 10% and increasing. The accuracy for non-spherical tumors was worse than for spherical tumors with same volume. Improvements in accuracy with partial volume compensation are expected.

Conclusions Reliable estimates of tumor activity can be obtained for tumors with sizes on the order of or larger than the system resolution. Quantifying activity in smaller tumors is challenging.

Research Support This work was supported by the Public Health Service Grant R01-CA109234