PET quantification: strategies for partial volume correction

Abstract

Despite the continuous evolution of positron emission tomography (PET) technology, the spatial resolution (SR) of PET images remains poor (4-6 mm) compared with that of computed tomography (CT) and magnetic resonance imaging (\1 mm). The reasons for this lie intrinsically in the physics behind PET and in the scanner design. This poor SR and the consequent partial volume effect (PVE), affecting image quality and the estimation of radioactivity concentration in small structures, complicate the anatomical localisation of focal radiotracer uptake signals and the application of a quantitative approach in the evaluation of PET studies. Methods for PVE compensation are therefore needed to improve the quality and quantitative accuracy of PET images. The simplest of the PVE compensation methods is the one based on recovery coefficients. Other techniques exploit high-resolution structural information extracted from co-registered CT or MR images to correct for the PVE. Others still are aimed instead at recovering the degradation effect by exploiting the point spread function of the PET system within the reconstruction process or using post-processing deconvolution algorithms. Regardless of the method employed, PVE compensation techniques generally assume motion-free PET data. In neurological studies, this assumption may be verified; in cardiological and oncological applications, on the other hand, organ/lesion motion and the PVE should necessarily be taken into account and corrected for sequentially or simultaneously. This paper deals with the PVE, its impact on qualitative and quantitative PET imaging, and possible approaches for PVE compensation.

Evaluation of image reconstruction algorithms encompassing Time-Of-Flight and Point Spread Function modelling for quantitative cardiac PET: Phantom studies

“…The results obtained in this work are expected to be applicable to most cardiac quantitative studies, independently from the tracer used, whether in perfusion studies (e.g., 15 O-water, 13 N-ammonia) or metabolic tracers studies, like 18 F-FDG or 11 C-acetate. However, the applicability might be limited for 82 Rb, since this nuclide has a much larger positron range, impacting PSF modelling.…”

“…It can be quantified with positron emission tomography (PET) and myocardial perfusion radiotracers such as 15 O-water, 13 Nammonia and 82 Rb. 3 New 18 F-tracers are also currently being evaluated, to expand the clinical use of this methodology.…”

“…14 The PSF modelling recovers the spatial resolution degradation and thus is able to reduce the associated partial volume effect (PVE), that is particularly relevant for structures with dimension \2.5 the full with at half of the maximum (FWHM) of the PET system. 15 The efficacy of OSEM with PSF modelling in cardiac 18 F-FDG studies was investigated by Le Meunier et al 16 who found that it improves contrast and defects definition while reducing image noise. Another study 11 also found that PSF reduces inter-voxel variance, as well as the amount of bias in low-statistics frames.…”

Classification and evaluation strategies of auto-segmentation approaches for PET: Report of AAPM task group No. 211

“…In principle, all the segmentation strategies not explicitly intended for blurred images, but widely used for imaging modalities less affected by PVE than PET (e.g., thresholding, region growing, gradient‐based algorithms, etc. ), can be applied to PET images after a PVE recovery step . PVE recovery can be performed after or during image reconstruction with algorithms taking into account a model of the scanner PSF .…”

“…), 103 can be applied to PET images after a PVE recovery step. 104 PVE recovery can be performed after [105][106][107][108][109] or during image reconstruction with algorithms taking into account a model of the scanner PSF. [110][111][112] These images, however, should be handled with caution since PVE recovery techniques can introduce artifacts (e.g., variance increase related to the Gibbs phenomenon).…”