Kjell Erlandsson

In vivo quantification of serotonin transporters using [11C]DASB and positron emission tomography in humans: modeling considerations

Positron emission tomography (PET) studies of the serotonin transporter (5-HTT) in the human brain are increasingly using the radioligand [11C]N, N-dimethyl-2-(2-amino-4-cyanophenylthio) benzylamine. A variety of models have been applied to such data in several published articles; however to date, these models have not been validated with test–retest data. We recruited 11 healthy subjects and conducted two identical scans on each subject on the same day. We considered four different models (one- and two-tissue compartment kinetic models, likelihood estimation in graphical analysis (LEGA; a bias-free alternative to the graphical method), and basis pursuit) along with fast noniterative approximations to the kinetic models. We considered four different outcome measures (total volume of distribution (VT), binding potential with (BP) and without (BP1), free-fraction adjustment, and specific-to-nonspecific equilibrium partition coefficient (BP2)). To assess the performance of each model, we compared results using six different metrics (percent difference (PD) and within-subject mean sum of squares for reproducibility, interclass coefficient for reliability, variance across subjects, identifiability based on bootstrap resampling of residuals for each method, and time stability analysis to determine minimal required scanning time). We considered analysis of both at the voxel level and at the region of interest (ROI) level and compared results from these two approaches to assess agreement. We determined that 100 mins of scanning time is adequate and that for ROI-level analysis, LEGA gives best results. Average PD is 5.51 for VT, 20.7 for BP, 17.2 for BP1, and 16.5 for BP2 across all regions. For voxel-level analysis we determined that the one-tissue compartment noniterative model is best.

An improved method for voxel-based partial volume correction in PET and SPECT

The limited spatial resolution of PET and SPECT leads to partial volume effects (PVE) that limit the quantification accuracy of these modalities. Partial volume correction (PVC) methods have been developed in the past utilizing high-resolution MRI images in combination with the known point-spread function (PSF) of the system. These methods can be broadly divided into voxel-based and volume of interest (VOI)-based methods. The voxel-based approach (Meltzer et al., JCAT, 1990, 14: 561–70) is based on segmentation of the MRI images into regions corresponding to gray matter (GM), white matter (WM) and cerebrospinal fluid. The correction is performed by assuming a uniform distribution within each region. A separate estimate of the WM value is required, and corrected values are obtained for voxels in the GM region only (target region). We have developed a new voxel-based PVC method in which a larger number of regions can be used, thereby reducing any bias introduced by the uniformity assumption. Also, with our new method, no estimate of the WM value is needed, and corrected values are obtained for each voxel in the whole image — not just one target region. Our new method (Multi-Target Correction (MTC)) involves an initial step using a VOI-based PVC method (Rousset et al. JNM, 1998, 39: 904–11). For the purpose of evaluation, simulated PET images were generated using a digital brain phantom, assuming a Gaussian PSF with 6 mm FWHM. The sensitivity of the method to various factors, such as errors in PSF determination, PET/MRI co-registration, and MRI segmentation, was determined. In addition, we examined the effects of tissue heterogeneity and noise. MTC was also applied to a PET study performed on a cylindrical phantom with 6 spherical inserts (diameters: 10, 12, 16, 20, 25, 32 mm, insert-to-background ratio 6) using an ECAT HR+ scanner (Siemens, Knoxville, TN, USA). MTC was performed using 7 VOIs. The simulations showed that MTC was reasonably robust with respect to errors in FWHM ( 2 mm), co-registration errors (<4.5 mm, or <6-), GM region dilation, as well as to region heterogeneity and noise. It was, however, quite sensitive to GM region erosion. Results from the phantom study are presented in the Fig. 1 below, showing that the whole image is corrected with improved contrast and definition of all the inserts. Further evaluations with human PET data are required before MTC can be used routinely.

Quantitative wavelet domain image processing of dynamic PET data

Neuroreceptor PET studies consisting of long dynamic data acquisitions result in data with low signal-to-noise ratio and limited spatial resolution. To address these problems we have developed a 3D wavelet-based image processing tool (wavelet filter, WF), containing both denoising and enhancement functionality. The filter is based on multi-scale thresholding and cross-scale regularization. These operations are data-driven, which may lead to non-linearity effects and hamper quantification of dynamic PET data. The aim of the present study was to investigate these effects using both phantom and human PET data. A phantom study was performed with a cylindrical phantom, filled with 18F, containing a number of spherical inserts filled with 11C. Human studies were performed on 9 healthy volunteers after injection of the serotonine transporter tracer [11C]DASB. Images from both phantom and human studies were reconstructed with filtered backprojection and post-processed by WF with a series of different denoising and enhancement parameter values. The phantom study was analyzed by computing the insert-to-background ratio as a function of time. The human study was analyzed with a 1-tissue compartment model for a series of brain regions. For the phantom study, linear relations were found between unprocessed and WF processed data for positive contrasts. However, for negative contrast, non-linearity effects were observed. For the human data, good correlation was obtained between results from unprocessed and WF processed data. Our results showed that, although non-linear effects may appear in low-contrast areas, it is possible to achieve accurate quantification with wavelet-based image processing

A new rebinning algorithm for 3D PET data

3D acquisition mode in PET is a way to increase the sensitivity of the scanner. To reconstruct the 3D data it is necessary to use either a fully 3D reconstruction algorithm or, alternatively, a rebinning algorithm followed by 2D reconstruction. The advantages of the 2nd approach are higher speed and reduced data size. Rebinning algorithms may become more important in the future with time-of-flight PET due to the increased dimensionality of the projection data. Although two exact rebinning algorithms have been developed, the most commonly used method today is the approximate Fourier rebinning algorithm. These three methods are based on analytic solutions in the frequency domain and involve geometric approximations, which may be significant for some scanner geometries. They also require attenuation corrected data, which can lead to complications since some reconstruction algorithms work better with non-attenuation corrected data. We have developed a novel rebinning algorithm, called reprojection rebinning (RP-RB), which is free from theoretical and geometric approximations as well as from the requirement of attenuation correction. It is based on an initial reconstruction of a 2D data-subset, which results in a somewhat sub-optimal utilization of the oblique data. The results can be improved, however, by an iterative procedure. We have tested RP-RB using simulated phantom data, showing an SD reduction in the central transaxial plane after the 1st, 2nd and 3rd iteration by a factor of 3.0, 3.5 and 4.1, respectively, as compared to 2D data only. There was also a very good spatial accuracy in images reconstructed from noiseless data, even for scanner geometries with large acceptance angles.