Abstract
1417
Objectives To validate the estimation of images of total distribution volume (VT) of [11C]harmine acquired with a HRRT scanner in the wavelet space to overcome the bias (~-40%) and variability of the voxel-wise quantification.
Methods 6 normal healthy subjects were scanned for 90 minutes after injection with [11C]harmine on a HRRT(Houle, BrainPET03). Arterial blood sampling and metabolite analysis were used to create a plasma input function. Regional time-activity curves were calculated and VTs were estimated using Logan graphical analysis(ROI-GA). Parametric images were generated in the wavelet space (Turkheimer, JCBF(20)2000) using dyadic wavelet transform(DWT-GA) with different levels of decomposition(J), filters length(k), and denoising filters. For comparison, graphical analysis was applied in a voxel-by-voxel basis(VOX-GA). Simulations were performed to study the feasibility for our data of other known strategies to overcome the bias introduced by noise in GA. The effects of the denoising filters on the resolution of the images were studied.
Results The optimal combination was a DWT-GA with J=4 and k=22 (Battle-Lemarie family) denoising the 2 finest resolution levels using a soft filter with an adaptive threshold. This approach shows no bias respect of ROI-GA independently of the VT value, increasing the FHWM of the scanner only a 10%.DWT-GA overcomes the 40% underestimation that VOX-GA produces. Our simulations show that other strategies to deal with the bias of GA without spatial constraints are not suitable for [11C]harmine and HRRT due to a high variability at this noise level.
Conclusions The present results indicate that solving the GA in the wavelets space is an accurate methodology for generation of parametric maps of [11C]harmine from the HRRT PET.
- © 2009 by Society of Nuclear Medicine