Comment on: 'A Poisson resampling method for simulating reduced counts in nuclear medicine images'

Area of Science:

• Nuclear Medicine Imaging
• Computational Physics
• Statistical Analysis

Background:

• Accurate simulation of reduced count images is crucial for data analysis in nuclear medicine.
• Previous methods, like Poisson resampling by White and Lawson, require experimental validation and comparison with alternatives.

Purpose of the Study:

• To reproduce and confirm the findings of White and Lawson regarding Poisson resampling for half-count images.
• To compare the performance of Poisson resampling with direct redrawing methods based on Poisson and Gaussian distributions.
• To evaluate the impact of rounding on counting statistics for different simulation methods.

Main Methods:

• Direct numerical simulation using Matlab.
• Investigation of Poisson resampling and two direct redrawing methods (Poisson and Gaussian distributions).
• Determination and comparison of statistical parameters (mean, standard deviation, skewness, excess kurtosis) of half-count/full-count ratios against theoretical Poisson values.

Main Results:

• Poisson resampling demonstrated superiority, accurately simulating statistical properties.
• Direct redrawing methods showed varying degrees of accuracy.
• Rounding off before saving significantly impacted counting statistics for low counts (<100), with Poisson resampling being unaffected.

Conclusions:

• Poisson resampling is the recommended method for simulating half-count images due to its robustness and accurate statistical representation.
• The method is particularly advantageous when dealing with rounding in image data, especially at low count levels.