Jeroen Verhaeghe
Ghent University
Network
Latest external collaboration on country level. Dive into details by clicking on the dots.
Publication
Featured researches published by Jeroen Verhaeghe.
Neuropsychopharmacology | 2013
Aryandokht Fotros; Kevin F. Casey; Kevin Larcher; Jeroen Verhaeghe; Sylvia M.L. Cox; Paul Gravel; Andrew J. Reader; Alain Dagher; Chawki Benkelfat; Marco Leyton
Drug-related cues are potent triggers for relapse in people with cocaine dependence. Dopamine (DA) release within a limbic network of striatum, amygdala and hippocampus has been implicated in animal studies, but in humans it has only been possible to measure effects in the striatum. The objective here was to measure drug cue-induced DA release in the amygdala and hippocampus using high-resolution PET with [18F]fallypride. Twelve cocaine-dependent volunteers (mean age: 39.6±8.0 years; years of cocaine use: 15.9±7.4) underwent two [18F]fallypride high-resolution research tomography–PET scans, one with exposure to neutral cues and one with cocaine cues. [18F]Fallypride non-displaceable-binding potential (BPND) values were derived for five regions of interest (ROI; amygdala, hippocampus, ventral limbic striatum, associative striatum, and sensorimotor striatum). Subjective responses to the cues were measured with visual analog scales and grouped using principal component analysis. Drug cue exposure significantly decreased BPND values in all five ROI in subjects who had a high-, but not low-, craving response (limbic striatum: p=0.019, associative striatum: p=0.008, sensorimotor striatum: p=0.004, amygdala: p=0.040, and right hippocampus: p=0.025). Individual differences in the cue-induced craving response predicted the magnitude of [18F]fallypride responses within the striatum (ventral limbic: r=0.581, p=0.048; associative: r=0.589, p=0.044; sensorimotor: r=0.675, p=0.016). To our knowledge this study provides the first evidence of drug cue-induced DA release in the amygdala and hippocampus in humans. The preferential induction of DA release among high-craving responders suggests that these aspects of the limbic reward network might contribute to drug-seeking behavior.
IEEE Transactions on Nuclear Science | 2007
Jeroen Verhaeghe; Yves D'Asseler; Steven Staelens; Stefaan Vandenberghe; Ignace Lemahieu
A maximum likelihood reconstruction algorithm for gated dynamic cardiac PET studies was developed and evaluated. A two dimensional tensor product spline basis spanning the time and gate domain is proposed. The activity variations introduced by the biochemical kinetics and cardiac motion are modeled as conic combinations of the B-spline basis functions. The basis explicitly takes the cyclic nature of the heart motion into account. We make use of the expectation-maximization (EM) algorithm to derive a closed form iteration scheme for the reconstruction. The proposed algorithm is validated through computer simulations of the dynamic NCAT beating heart phantom and the kinetics of 13N-ammonia uptake. We used the Monte Carlo simulator GATE to simulate a typical PET scanner. We qualitatively found that a reconstruction using cubic splines resulted in smoother images while better delineating the myocardial wall. Fourth order spline modeling reduced the mean squared error (MSE) of binned reconstruction by 56% whereas conventional Gaussian filtering reduced the MSE by 38%. Spline modeling in the gate domain reduced the MSE by 72% compared to a reduction of 47% obtained with filtering. Quantitative evaluation of the reconstructed motion information suggested that the number of basis functions in the gate domain could be reduced from 16 in a framed approach to 5 for reconstructions using a higher order spline interpolation
Medical Physics | 2007
Jeroen Verhaeghe; Yves D'Asseler; Stefaan Vandenberghe; Steven Staelens; Ignace Lemahieu
The use of a temporal B-spline basis for the reconstruction of dynamic positron emission tomography data was investigated. Maximum likelihood (ML) reconstructions using an expectation maximization framework and maximum A-posteriori (MAP) reconstructions using the generalized expectation maximization framework were evaluated. Different parameters of the B-spline basis of such as order, number of basis functions and knot placing were investigated in a reconstruction task using simulated dynamic list-mode data. We found that a higher order basis reduced both the bias and variance. Using a higher number of basis functions in the modeling of the time activity curves (TACs) allowed the algorithm to model faster changes of the TACs, however, the TACs became noisier. We have compared ML, Gaussian postsmoothed ML and MAP reconstructions. The noise level in the ML reconstructions was controlled by varying the number of basis functions. The MAP algorithm penalized the integrated squared curvature of the reconstructed TAC. The postsmoothed ML was always outperformed in terms of bias and variance properties by the MAP and ML reconstructions. A simple adaptive knot placing strategy was also developed and evaluated. It is based on an arc length redistribution scheme during the reconstruction. The free knot reconstruction allowed a more accurate reconstruction while reducing the noise level especially for fast changing TACs such as blood input functions. Limiting the number of temporal basis functions combined with the adaptive knot placing strategy is in this case advantageous for regularization purposes when compared to the other regularization techniques.
Current Alzheimer Research | 2015
Ann-Marie Waldron; Leonie wyffels; Jeroen Verhaeghe; Astrid Bottelbergs; Jill C. Richardson; Jonathan Kelley; Mark Schmidt; Sigrid Stroobants; Xavier Langlois; Steven Staelens
Positron emission tomography studies of cerebral glucose utilization and amyloid-β deposition with fluoro-deoxy-D-glucose ([(18)F]-FDG) and amyloid tracers have shown characteristic pathological changes in Alzheimers Disease that can be used for disease diagnosis and monitoring. Application of this technology to preclinical research with transgenic animal models would greatly facilitate drug discovery and further understanding of disease processes. The results from preclinical studies with these imaging biomarkers have however been highly inconsistent, causing doubts over whether animal models can truly replicate an AD-like phenotype. In this study we performed in vivo imaging with [(18)F]-FDG and [(18)F]-AV45 in double transgenic TASTPM mice, a transgenic model that been previously demonstrated high levels of fibrillar amyloid-β and decreases in cerebral glucose utilization with ex vivo techniques. Our results show widespread and significant retention of [(18)F]-AV45 (p < 0.0001) in aged TASTPM mice in addition to significant regional decreases in [(18)F]-FDG uptake (p < 0.05). In vivo quantification of amyloid-β showed a strong (Pearsons r = 0.7078), but not significant (p = 0.1156), positive correlation with ex vivo measures suggesting some limitations on tracer sensitivity. In the case of [(18)F]-FDG, voxelwise analysis greatly enhanced detection of hypometabolic regions. We further evidenced modest neuronal loss (thalamus p = 0.0318) that could underlie the observed hypometabolism. This research was performed in conjunction with the European Communitys Seventh Framework Program (FP7/2007-2013) for the Innovative Medicine Initiative under the PharmaCog Grant Agreement no.115009.
ieee nuclear science symposium | 2007
Stefaan Vandenberghe; Jeroen Verhaeghe; Ignace Lemahieu; Samuel Matej; Margaret E. Daube-Witherspoon; Joel S. Karp; Michel Guerchaft; Anne Bol; L. van Elmbt
TOF resolution is known to degrade with increasing count rate. During TOF-PET reconstruction the timing resolution of the data is used as an input for the reconstruction algorithm. The effect of the kernel width on the reconstruction was investigated in this paper. We looked at contrast recovery and background uniformity. Noise free simulation data were used together with high count measured TOF-PET data. The results clearly indicate that only the correct kernel should be used to reconstruct the data. Other kernels result in wrong contrast and less uniform background regions. Therefore it would be very useful to be able to estimate the TOF kernel from the data themselves. It is first shown that the likelihood reaches a maximum at the correct timing resolution. Then a method to estimate the kernel from both a non-TOF reconstruction and the measured TOF data is described. The determination of the timing resolution is performed with an iterative method using the non-TOF MLEM reconstruction and Richardson-Lucy deconvolution.
Journal of Alzheimer's Disease | 2017
Ellis Niemantsverdriet; Julie Ottoy; Charisse Somers; Ellen Elisa De Roeck; Hanne Struyfs; Femke Soetewey; Jeroen Verhaeghe; Tobi Van den Bossche; Sara Van Mossevelde; Johan Goeman; Peter Paul De Deyn; Peter Mariën; Jan Versijpt; Kristel Sleegers; Christine Van Broeckhoven; Leonie wyffels; Adrien Albert; Sarah Ceyssens; Sigrid Stroobants; Steven Staelens; Maria Bjerke; Sebastiaan Engelborghs
Background: Evidence suggests that the concordance between amyloid-PET and cerebrospinal fluid (CSF) amyloid-β (Aβ) increases when the CSF Aβ1–42/Aβ1–40 ratio is used as compared to CSF Aβ1–42 levels alone. Objective: In order to test this hypothesis, we set up a prospective longitudinal study comparing the concordance between different amyloid biomarkers for Alzheimer’s disease (AD) in a clinical setting. Methods: Seventy-eight subjects (AD dementia (n = 17), mild cognitive impairment (MCI, n = 48), and cognitively healthy controls (n = 13)) underwent a [18F]Florbetapir ([18F]AV45) PET scan, [18F]FDG PET scan, MRI scan, and an extensive neuropsychological examination. In a large subset (n = 67), a lumbar puncture was performed and AD biomarkers were analyzed (Aβ1–42, Aβ1–40, T-tau, P-tau181). Results: We detected an increased concordance in the visual and quantitative (standardized uptake value ratio (SUVR) and total volume of distribution (VT)) [18F]AV45 PET measures when the CSF Aβ1–42/Aβ1–40 was applied compared to Aβ1–42 alone. CSF biomarkers were stronger associated to [18F]AV45 PET for SUVR values when considering the total brain white matter as reference region instead of cerebellar grey matter Conclusions: The concordance between CSF Aβ and [18F]AV45 PET increases when the CSF Aβ1–42/Aβ1–40 ratio is applied. This finding is of most importance for the biomarker-based diagnosis of AD as well as for selection of subjects for clinical trials with potential disease-modifying therapies for AD.
Journal of Alzheimer's Disease | 2016
Ann-Marie Waldron; Leonie wyffels; Jeroen Verhaeghe; Jill C. Richardson; Mark Schmidt; Sigrid Stroobants; Xavier Langlois; Steven Staelens
We aimed to monitor the timing of amyloid-β deposition in relation to changes in brain function using in vivo imaging with [18F]-AV45 and [18F]-FDG in a mouse model of Alzheimer’s disease. TASTPM transgenic mice and wild-type controls were scanned longitudinally with [18F]-AV45 and [18F]-FDG before (3 months of age) and at multiple time points after the onset of amyloid deposition (6, 9, 12, and 15 months of age). As expected with increasing amyloidosis, TASTPM mice demonstrated progressive age-dependent increases in [18F]-AV45 uptake that were significantly higher than for WT from 9 months onwards and correlated to ex vivo measures of amyloid burden. The metabolism of [18F]-AV45 produces several brain penetrant radiometabolites and normalization to a reference region helps to negate this non-specific binding and improve the sensitivity of [18F]-AV45. The observed trajectory of [18F]-FDG alterations deviated from our proposed hypothesis of gradual decreases with worsening amyloidosis. While [18F]-FDG uptake in TASTPM mice was significantly lower than that of WT at 9 months, reduced [18F]-FDG was not associated with aging in TASTPM mice. Moreover, [18F]-FDG uptake did not correlate to measures of ex vivo amyloid burden. Our findings suggest that while amyloid-β is sufficient to induce hypometabolism, these pathologies are not linked in a dose-dependent manner in TASTPM mice.
nuclear science symposium and medical imaging conference | 2012
Etienne Létourneau; Jeroen Verhaeghe; Andrew J. Reader
This work is a task-oriented quantitative analysis to assess the best range of reconstruction algorithms parameters which are most suited for quantitative PET imaging of the brain. Starting from a general iterative weighted-least squares objective function, well-known methods such as the Maximum Likelihood Expectation Maximization (MLEM), the Image Space Reconstruction Algorithm (lSRA) and related techniques are obtained. A detailed analysis was done for single-frame imaging as well as a comprehensive study on the impact of post-smoothing by applying a Gaussian convolution kernel at each and every possible post-reconstruction iteration number in order to achieve the lowest mean absolute error (MAE). It has been clearly demonstrated that an appropriate choice of iteration number and post-smoothing level makes most algorithms deliver similar MAE within an interval of 5%, with a considerable reduction of the MAE in comparison to non-optimized reconstructions. Also, reconstruction method performance was influenced by the total number of counts and the activity distribution, meaning that each radiopharmaceutical is not always best reconstructed by the same method. Broadly speaking, MLEM with a point spread function (PSF), ISRA PSF with smoothed expected data as weighting factors and filtered backprojection (FBP), when applied with a suitable level of post-smoothing, were the methods offering the best quantitative images and from these three methods, FBP offered the most robust performance for a broad range of post-smoothing levels. An appropriate choice of iteration number and post-smoothing level for a task-oriented analysis significantly lowered the MAE in comparison with the standard practice of using default fixed numbers of iterations and/or post-smoothing.
ieee nuclear science symposium | 2006
Yu Chen; Stefaan Vandenberghe; Steven Staelens; Jeroen Verhaeghe; Stephen J. Glick
The spatially variant detector response can reduce spatial resolution in PET imaging. In iterative reconstruction methods, the detector response can be modeled into the system response matrix (SRM). Unfortunately, the SRM for current PET scanners can be very large. We have been evaluating PET reconstruction using generalized natural pixel (GNP) functions. With these pixel functions, the SRM becomes block-circulant for a ring-PET scanner, thereby substantially reducing the number of non-redundant elements in the SRM. Application of the generalized natural pixel functions assumes a perfect rotationally symmetric system. There are no such PET scanners in reality. We developed a method to correct and match an actual PET scanner geometry to a virtual rotationally symmetric system. The Geant4 based GATE Monte Carlo simulation code was used to model a 2D version of the Philips Allegro PET scanner consisting of one ring with 616 GSO crystals. The simulation code modeled all interactions within the detector. To evaluate and compare reconstruction algorithms, a simulated 2D phantom and two physical phantoms were used to collect simulated or experimental LOR-binned fan beam sinogram data. Reconstruction was performed using either an algebraic reconstruction technique (ART) with generalized natural pixels, or a LOR-based maximum-likelihood expectation maximization algorithm (MLEM) using Siddon ray tracing. We studied the contrast versus noise as a function of iteration number. The GNP-ART method clearly outperforms the LOR-MLEM with higher contrast at the same noise level. This can be simply attributed to the improvement of spatial resolution in GNP-ART by modeling spatially variant detector response in the SRM.
ieee nuclear science symposium | 2006
Jeroen Verhaeghe; Ronald Phlypo; Stefaan Vandenberghe; Steven Staelens; Yves D'Asseler; Ignace Lemahieu
In this paper we try to find an optimal temporal basis to reconstruct dynamic list-mode PET data. The goal of this method is to avoid over-fitting of and to avoid the introduction of too much bias in the reconstructed Time Activity Curves (TACs). The optimal basis is estimated from the list-mode data during reconstruction. In particular we have evaluated the Akaike Information Criterion (AIC) as objective function for the determination of the optimal number of temporal B-spline basis functions. The optimal number is spatially variant and is determined for different physiological regions in the image. The required segmentation based on the image sequences clusters the pixels of different physiological regions. We have evaluated different clustering methods. Numerical experiments of simulated cardiac PET data showed that the method based on an initial Non-negative Matrix Factorization (NMF) which includes a Poissonian noise model performed the best. Our results show that the AIC based determination of the optimal number of basis functions was useful for dynamic PET reconstruction, especially for an accurate reconstruction of the blood input curves.