Optimizing 4D-PET/CT imaging for heterogeneity quantification by texture features

The number of time-bins (TB) in retrospectively gated (4D-) PET/CT should be a trade-off between optimizing statistics and minimizing motion effect. The aim of this work was to investigate the impact of TB in heterogeneity quantification by PET texture features (TF) for lesions affected by respiratory motion. Five heterogeneous lesions were simulated by a mixture of radio-compound and alginate. Realistic 2D breathing pattern was applied to the phantoms. PET/CT acquisition (10 min/ 36 s) was retrospectively gated using different TB (4, 6, 8 and 10). Static (S-) PET/CT acquisition (15 min/ 3.7 s) was also performed when no movement was applied to the phantoms. Our analysis involved 11 metrics for each of the 5 heterogeneous phantoms (55 data series). Metrics were 3 conventional indexes and 8 TF. Impact of TB in PET quantification was evaluated in terms of metrics variability along TB and linear correlation with respect to values derived from ideal contours on S-image, reference values (M
_ref
). From the comparison between different TB, significant worse variability was obtained for TB=10: coefficient of variance along TB averaged over the 11 metrics was COV
_mean
=16±13 and 11/55 data metrics did not follow a normal distribution along TB. Non significant differences were observed for the variability obtained for TB=4, 6 and 8. The strongest correlation with respect to M
_ref
was observed for TB=6. From our results, TB<;10 should be preferred for PET quantification of lesions affected by respiratory movement. Although TB=6 could be suggested as optimal 4D-PET/CT, future validation with larger sample of heterogeneous inserts and movements is recommended.