The impact of reconstruction algorithms and time of flight information on PET/CT image quality

Abstract

Background. The aim of the study was to explore the influence of various time-of-flight (TOF) and non-TOF reconstruction algorithms on PET/CT image quality.

Materials and methods. Measurements were performed with a triple line source phantom, consisting of capillaries with internal diameter of ~ 1 mm and standard Jaszczak phantom. Each of the data sets was reconstructed using analytical Filtered Back Projection (FBP) algorithm, Iterative Ordered Subsets Expectation Maximization (OSEM) algorithm (4 iterations, 24 subsets) and Iterative True-X algorithm incorporating a specific point spread function (PSF) correction (4 iterations, 21 subsets). Baseline OSEM (2 iterations, 8 subsets) was included for comparison. Procedures were undertaken following the NEMA NU-2-2001 protocol.

Results. Measurement of spatial resolution in FWHM was 5.2 mm, 4.5 mm and 2.9 mm for FBP, OSEM and True-X; and 5.1 mm, 4.5 mm and 2.9 mm for FBP+TOF, OSEM+TOF and True-X+TOF respectively. Assessment of reconstructed Jaszczak images at different concentration ratios showed that incorporation of TOF information improves cold contrast, while hot contrast only slightly, however the most prominent improvement could be seen in background variability - noise reduction.

Conclusions. On the basis of the results of investigation we concluded, that incorporation of TOF information in reconstruction algorithm mostly affects reduction of the background variability (levels of noise in the image), while the improvement of spatial resolution due to incorporation of TOF information is negligible. Comparison of traditional and modern reconstruction algorithms showed that analytical FBP yields comparable results in some parameter measurements, such as cold contrast and relative count error. Iterative methods show highest levels of hot contrast, when TOF and PSF corrections were applied simultaneously.

Key words: time of flight, PET/CT, point spread function, reconstruction algorithm, image quality