Background
[11C]Diprenorphine (DPN) PET images all subtypes of opioid receptors. Various methods of data acquisition and analysis are used. We describe our revisited methodology for acquiring high quality DPN brain PET studies and compare the influence of various combinations of movement correction, different amounts of injected activity and derivation of input function on test-retest data.
Methods
Ten healthy controls were studied twice. Five each were injected with ∼135 MBq and ∼185 MBq of [11C]DPN. All had high resolution MRI and quantitative 95' listmode [11C]DPN PET, rebinned into 32 time frames, on a Siemens/CTI ECAT HR++ scanner. Movement was assessed on decay-corrected time-activity curves derived from ROIs drawn on summed activity (ADD) images. Movement correction was performed by wavelet denoising the dynamic time frames, coregistering them to the first 120 second frame with a mutual information method, and reslicing the original time series (MVCORR images). Online blood collection was 90' (∼135 MBq group) and shortened to 15' for the ∼185 MBq group. Both groups had discrete blood samples throughout the length of the scan for calibration and metabolite analysis. Metabolite-corrected arterial plasma input functions were derived using only the first 15' of online blood collection for both groups; and using all 90' of data for the ∼135 MBq group (“CONTINUOUS_IF”). Spectral analysis was used to produce parametric images of DPN volume-of-distribution (VD). In addition, 28 minutes to 58 minutes ADD images were created from the MVCORR images for the ∼185 MBq group. This led to four sets of VD images (∼135 MBq group and ∼185 MBq group, both MVCORR and Non-MVCORR), one set of ∼185 MBq MVCORR ADD images, and one set of ∼135 MBq MVCORR CONTINUOUS_IF images. For image sampling, our probabilistic brain atlas was warped onto each individual's MRI scan using the deformations toolbox in the Statistical Parametric Mapping software (SPM2) and these were then coregistered onto each individual PET. We assessed hippocampus, thalamus, cerebellum and inferior frontal gyrus. Test-retest bias (order effect) and significance of differences in test-retest reliability were assessed using MANOVA. Test-retest reliability was calculated using intraclass correlation coefficients (ICC).
Results
There was a small but significant order effect, with the second scan showing a VD on average 1.8% lower. The rank order of ICCs was ∼185 MBq Non-MVCORR (0.949) > ∼185 MBq MVCORR (0.944) > ∼185 MBq MVCORR ADD (0.934) > ∼135 MBq Non-MVCORR (0.876) > ∼135 MBq MVCORR CONTINUOUS_IF (0.863) > ∼135 MBq MVCORR (0.782). The higher noise level of the ∼135 MBq images was captured by MANOVA with borderline significance (p<0.069); there was a significant improvement for the ∼185 MBq group in the inferior frontal gyrus, right thalamus and right cerebellum. There was no overall effect of MVCORR. However, in one subject with significant movement, correction vastly improved test-retest reliability, approximately halving the difference between first and second datapoint.
Summary and Conclusions
Overall, injecting the higher dose of ∼185 MBq markedly improved test-retest reliability, with % T-RT variation decreasing by ∼20–50%. MVCORR slightly worsened test-retest reliability in the ∼135 MBq group. It had no overall effect in the ∼185 MBq group but appears indicated in those subjects in whom movement is detected.