Comparison of the different reconstruction algorithms for Philips GEMINI PET/CT / 中国医学影像技术

ABSTRACT

Objective To evaluate the effects of different reconstruction algorithms on image quality for Philips GEMINI PET/CT. Methods Jaszczak phantom were scanned on the GEMINI PET/CT system, and the raw data were reconstructed using filtered-back projection with Hanning filter (FBP-Hanning), filtered-back projection with Butterworth filter (FBP-Butterworth), ordered subset expectation maximization (OSEM), row-action maximum likelihood algorithm (RAMLA) and three-dimensional row-action maximum likelihood algorithm (3D-RAMLA), respectively. The resolution, uniformity, contrast of images and the time of different reconstruction algorithms were compared. Results The reconstruction time was 180 s, 130 s, 120 s, 85 s and 80 s for 3D-RAMLA, RAMLA, OSEM, FBP-Hanning and FBP-Butterworth respectively in phantom studies. The smallest rods with diameter of 4.8 mm of the phantom could be observed for FBP- Butterworth and 6.4 mm for other algorithms in case of high counts. The image contrast of 3D-RAMLA were better than that other algorithms, and the image uniformity of 3D-RAMLA and RAMLA were better than those of other algorithms. The resolution, uniformity and contrast of images with all algorithms decreasd in case of low counts, and the image quality of FBP-Butterworth was not good enough for clinical studies. Conclusion Image quality is variable with different reconstruction algorithms. For clinical PET imaging, it is necessary to choose proper algorithms.