Easy blur estimation in PET images including motion corrupted edges

The PET image is often described as blurred, however, it is our knowledge that very few studies are focused on the evaluation of blurring in clinical images. This paper describes an easy and practical method for global blurring estimation of PET images based on the analysis of the objects edge. This blur index is associated with the full width at half maximum (fwhm) of edges response to first-order directional Gaussian derivative operator. Two types of edges are considered: in addition to the usually considered edges corrupted by Gaussian filtering, an estimate of corrupted edges by motion is also included. The procedure verification is carried out in different radioactive environments with different signal-to-background ratio (SIB) and applying post-processed Gaussian filters with kernel sizes from 3 mm to 9 mm and through a mobile platform that induces a simple harmonic motion of different amplitudes from 3 mm to 15 mm. The results show that the proposed procedure is able to distinguish the induced blurring, either due to filtering or motion applied. It is robust to parameters selection and it could be applied to clinical images. It is also possible to provide some insight about direction and magnitude of motion from data collected at each angle. Examples of the application with clinical and non-clinical images are shown.