Optimization of iterative reconstruction parameters with 3-dimensional resolution recovery, scatter and attenuation...

The aim of this study was to determine the optimal reconstruction parameters on ordered-subset expectation maximization iterative reconstruction with resolution recovery, scatter and attenuation correction (OSEM(RRSCAC)) on (123)I-FP-CIT SPECT in terms of the image quality, quantification and diagnostic ability.

We evaluated the quality and quantification in (123)I-FP-CIT SPECT images obtained by different reconstruction parameters using the anthropomorphic striatal phantom. The phantom images were acquired using a SPECT/CT system equipped with a low- and medium-energy general-purpose collimator and then were reconstructed using OSEM(RRSCAC) with various update numbers and the full width at half maximum (FWHM) of the Gaussian filter. The count ratio of the striatum, the coefficient of variation (CV) on the background and the specific binding ratio (SBR) were calculated to determine the optimal reconstruction parameters. Then, 42 consecutive patients who underwent (123)I-FP-CIT SPECT were selected for clinical study. The patients were grouped into potentially decreasing group (20 with Parkinson's disease, 5 with parkinsonian syndrome and 5 with dementia with Lewy bodies) and potentially normal binding group (4 with Alzheimer disease and 8 with essential tremor). Clinical images were reconstructed using OSEM(RRSCAC) and FBP with Chang's AC (FBP(AC)). The performance of OSEM(RRSCAC) with the optimal reconstruction parameters was evaluated according to the receiver operating characteristic analysis and visual assessment.

The count ratio increased with an increasing update number and converged uniformly with an update number over 90. The CV continuously increased with the update number. The SBR also increased with an increasing update number and converged uniformly with an update number of 90 or larger. The FWHM of the Gaussian filter influenced the image quality and quantification. In the clinical study, the area under curve (AUC) for the OSEM(RRSCAC) and FBP(AC) were 0.9812 and 0.9759, respectively. The quality of the SPECT images of OSEM(RRSCAC) (3.8 ± 0.8) was significantly superior to that of FBP(AC) (2.2 ± 0.7).

We determined that the optimal reconstruction parameter for OSEM(RRSCAC) was 90 update numbers with 6.6 mm FWHM of the Gaussian filter. Our results suggested that the optimal reconstruction parameters have a potential to improve the performance and the image quality of (123)I-FP-CIT SPECT in comparison with the FBP reconstruction.