A Poisson Resampling Method for Simulating Reduced Counts in Nuclear Medicine Images

Nuclear medicine computers now commonly offer resolution recovery and other software techniques which have been developed to improve image quality for images with low counts. These techniques potentially mean that these images can give equivalent clinical information to a full-count image. Reducing the number of counts in nuclear medicine images has the benefits of either allowing reduced activity to be administered or reducing acquisition times. However, because acquisition and processing parameters vary, each user should ideally evaluate the use of images with reduced counts within their own department, and this is best done by simulating reduced-count images from the original data. Reducing the counts in an image by division and rounding off to the nearest integer value, even if additional Poisson noise is added, is inadequate because it gives incorrect counting statistics. This technical note describes how, by applying Poisson resampling to the original raw data, simulated reduced-count images can be obtained while maintaining appropriate counting statistics. The authors have developed manufacturer independent software that can retrospectively generate simulated data with reduced counts from any acquired nuclear medicine image.