Validation of the count-reduction method for planar bone scintigraphy: a phantom study focused on hot-lesion detection

Poisson resampling can be used to synthesize low-count nuclear medicine images from full-count data. However, its fidelity relative to real short acquisitions remains uncertain. Using planar acquisitions of JSP and SIM2 bone phantoms filled with 99mTc, we compared the Poisson-generated and measured images at matched-count reductions (10–90%). The simulated total counts matched the measured counts at all reduction levels. The NMSE increased with a greater count reduction, and the values were similar between the groups at each level. In the bone phantom, the contrast ratio and CNR declined with fewer counts, whereas the %CV increased; the trends and values were similar between the simulated and measured images, even at a 90% reduction rate. These findings indicate that the Poisson-based count reduction can reproduce, without apparent systematic bias, the statistics and image-quality characteristics of true low-count images, supporting the safe and efficient optimization of low-dose and short-time protocols without additional patient scans.