Click here to


Are you sure ?

Yes, do it No, cancel

A New Methodology for Reconstruction of 4D-PET Images

H Zhong*, X Li , Medical College of Wisconsin, Milwaukee, WI


(Sunday, 7/14/2019) 4:00 PM - 6:00 PM

Room: 304ABC

Purpose: To develop a novel iterative image-to-image (III) reconstruction method to generate 4D-PET images without using any target-tracking device for PET data acquisition.

Methods: A free-breathing (FB) PET image was considered as a weighted average of PET images acquired at different respiration phases. PET images at each phase were assumed to be deformed from a virtual reference image. To reconstruct the reference image, deformation maps were used to establish a correlation between the reference image and the FB-PET image, and this correlation was reformulated as a set of composition equations. Two numerical algorithms, a generalized minimal residual (GMRES) algorithm and a maximum-likelihood estimation-maximization (MLEM) algorithm, were implemented to solve these equations. The implementations were validated using three computational phantoms and demonstrated on a patient’s FB-PET dataset. The deformation maps for the patient data were derived by B-Spline-based registrations performed on the patient’s 4DCT. Universal-quality-index (UQI), signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) were calculated to evaluate the quality of the reconstructed images.

Results: For the three computational phantoms, their reference images were exactly recovered by the two numerical algorithms, and UQIs were increased from 0.85, 0.87 and 0.64 for the three blurred phantom images to 1 for their reconstructed reference images. When the blurred phantom images and their deformation maps were perturbed with each by 10%, the minimal SNR of their reconstructed images for the two types of perturbations was 4.4 and 1.4, respectively. For the patient’s FB-PET, the III-reconstruction increased CNR, SUVmax and SUVpeak by 63.7%, 62.3% and 23.3%, respectively.

Conclusion: A novel image-to-image reconstruction method has been developed to generate 4D-PET, removing respiratory motion-induced artifacts in FB-PET images. This method does not require sorting PET projection data, and therefore implementation of this method will help simplify clinical procedures and reduce both equipment cost and data acquisition time.


Deformation, Reconstruction, Quantitative Imaging


IM- PET : Quantitative imaging/analysis

Contact Email