Positron emission tomography partial volume correction: estimation and algorithms

Partial volume correction in longitudinal tau PET studies: is it really needed?

BACKGROUND

[18F]flortaucipir (FTP) tau PET quantification is known to be affected by non-specific binding in off-target regions. Although partial volume correction (PVC) techniques partially account for this effect, their inclusion may also introduce noise and variability into the quantification process. While the impact of these effects has been studied in cross-sectional designs, the benefits and drawbacks of PVC on longitudinal FTP studies is still under scrutiny. The aim of this work was to study the performance of the most common PVC techniques for longitudinal FTP imaging.

METHODS

A cohort of 247 individuals from the Alzheimer's Disease Neuroimaging Initiative with concurrent baseline FTP-PET, amyloid-beta (Aβ) PET and structural MRI, as well as with follow-up FTP-PET and MRI were included in the study. FTP-PET scans were corrected for partial volume effects using Meltzer's, a simple and popular analytical PVC, and both the region-based voxel-wise (RBV) and the iterative Yang (iY) corrections. FTP SUVR values and their longitudinal rates of change were calculated for regions of interest (ROI) corresponding to Braak Areas I-VI, for a temporal meta-ROI and for regions typically displaying off-target FTP binding (caudate, putamen, pallidum, thalamus, choroid plexus, hemispheric white matter, cerebellar white matter, and cerebrospinal fluid). The longitudinal correlation between binding in off-target and target ROIs was analysed for the different PVCs. Additionally, group differences in longitudinal FTP SUVR rates of change between Aβ-negative (A-) and Aβ-positive (A+), and between cognitively unimpaired (CU) and cognitively impaired (CI) individuals, were studied. Finally, we compared the ability of different partial-volume-corrected baseline FTP SUVRs to predict longitudinal brain atrophy and cognitive decline.

RESULTS

Among off-target ROIs, hemispheric white matter showed the highest correlation with longitudinal FTP SUVR rates from cortical target ROIs (R2=0.28-0.82), with CSF coming in second (R2=0.28-0.42). Application of voxel-wise PVC techniques minimized this correlation, with RBV performing best (R2=0.00-0.07 for hemispheric white matter). PVC also increased group differences between CU and CI individuals in FTP SUVR rates of change across all target regions, with RBV again performing best (No PVC: Cohen's d = 0.26-0.66; RBV: Cohen's d = 0.43-0.74). These improvements were not observed for differentiating A- from A+ groups. Additionally, voxel-wise PVC techniques strengthened the correlation between baseline FTP SUVR and longitudinal grey matter atrophy and cognitive decline.

CONCLUSION

Quantification of longitudinal FTP SUVR rates of change is affected by signal from off-target regions, especially the hemispheric white matter and the CSF. Voxel-wise PVC techniques significantly reduce this effect. PVC provided a significant but modest benefit for tasks involving the measurement of group-level longitudinal differences. These findings are particularly relevant for the estimations of sample sizes and analysis methodologies of longitudinal group studies.

Can Internal Carotid Arteries Be Used for Noninvasive Quantification of Brain PET Studies?

Because of the limited axial field of view of conventional PET scanners, the internal carotid arteries are commonly used to obtain an image-derived input function (IDIF) in quantitative brain PET. However, time-activity curves extracted from the internal carotids are prone to partial-volume effects due to the limited PET resolution. This study aimed to assess the use of the internal carotids for quantifying brain glucose metabolism before and after partial-volume correction. Methods: Dynamic [18F]FDG images were acquired on a 106-cm-long PET scanner, and quantification was performed with a 2-tissue-compartment model and Patlak analysis using an IDIF extracted from the internal carotids. An IDIF extracted from the ascending aorta was used as ground truth. Results: The internal carotid IDIF underestimated the area under the curve by 37% compared with the ascending aorta IDIF, leading to Ki values approximately 17% higher. After partial-volume correction, the mean relative Ki differences calculated with the ascending aorta and internal carotid IDIFs dropped to 7.5% and 0.05%, when using a 2-tissue-compartment model and Patlak analysis, respectively. However, microparameters (K 1, k 2, k 3) derived from the corrected internal carotid curve differed significantly from those obtained using the ascending aorta. Conclusion: These results suggest that partial-volume-corrected internal carotids may be used to estimate Ki but not kinetic microparameters. Further validation in a larger patient cohort with more variable kinetics is needed for more definitive conclusions.

A cycle-consistent adversarial network for brain PET partial volume correction without prior anatomical information

PURPOSE

Partial volume effect (PVE) is a consequence of the limited spatial resolution of PET scanners. PVE can cause the intensity values of a particular voxel to be underestimated or overestimated due to the effect of surrounding tracer uptake. We propose a novel partial volume correction (PVC) technique to overcome the adverse effects of PVE on PET images.

METHODS

Two hundred and twelve clinical brain PET scans, including 50 18F-Fluorodeoxyglucose (18F-FDG), 50 18F-Flortaucipir, 36 18F-Flutemetamol, and 76 18F-FluoroDOPA, and their corresponding T1-weighted MR images were enrolled in this study. The Iterative Yang technique was used for PVC as a reference or surrogate of the ground truth for evaluation. A cycle-consistent adversarial network (CycleGAN) was trained to directly map non-PVC PET images to PVC PET images. Quantitative analysis using various metrics, including structural similarity index (SSIM), root mean squared error (RMSE), and peak signal-to-noise ratio (PSNR), was performed. Furthermore, voxel-wise and region-wise-based correlations of activity concentration between the predicted and reference images were evaluated through joint histogram and Bland and Altman analysis. In addition, radiomic analysis was performed by calculating 20 radiomic features within 83 brain regions. Finally, a voxel-wise two-sample t-test was used to compare the predicted PVC PET images with reference PVC images for each radiotracer.

RESULTS

The Bland and Altman analysis showed the largest and smallest variance for 18F-FDG (95% CI: - 0.29, + 0.33 SUV, mean = 0.02 SUV) and 18F-Flutemetamol (95% CI: - 0.26, + 0.24 SUV, mean =  - 0.01 SUV), respectively. The PSNR was lowest (29.64 ± 1.13 dB) for 18F-FDG and highest (36.01 ± 3.26 dB) for 18F-Flutemetamol. The smallest and largest SSIM were achieved for 18F-FDG (0.93 ± 0.01) and 18F-Flutemetamol (0.97 ± 0.01), respectively. The average relative error for the kurtosis radiomic feature was 3.32%, 9.39%, 4.17%, and 4.55%, while it was 4.74%, 8.80%, 7.27%, and 6.81% for NGLDM_contrast feature for 18F-Flutemetamol, 18F-FluoroDOPA, 18F-FDG, and 18F-Flortaucipir, respectively.

CONCLUSION

An end-to-end CycleGAN PVC method was developed and evaluated. Our model generates PVC images from the original non-PVC PET images without requiring additional anatomical information, such as MRI or CT. Our model eliminates the need for accurate registration or segmentation or PET scanner system response characterization. In addition, no assumptions regarding anatomical structure size, homogeneity, boundary, or background level are required.

Image-derived input functions from dynamic 15O-water PET scans using penalised reconstruction

BACKGROUND

Quantitative positron emission tomography (PET) scans of the brain typically require arterial blood sampling but this is complicated and logistically challenging. One solution to remove the need for arterial blood sampling is the use of image-derived input functions (IDIFs). Obtaining accurate IDIFs, however, has proved to be challenging, mainly due to the limited resolution of PET. Here, we employ penalised reconstruction alongside iterative thresholding methods and simple partial volume correction methods to produce IDIFs from a single PET scan, and subsequently, compare these to blood-sampled input curves (BSIFs) as ground truth. Retrospectively we used data from sixteen subjects with two dynamic ¹⁵O-labelled water PET scans and continuous arterial blood sampling: one baseline scan and another post-administration of acetazolamide.

RESULTS

IDIFs and BSIFs agreed well in terms of the area under the curve of input curves when comparing peaks, tails and peak-to-tail ratios with R² values of 0.95, 0.70 and 0.76, respectively. Grey matter cerebral blood flow (CBF) values showed good agreement with an average difference between the BSIF and IDIF CBF values of 2% ± and a coefficient of variation (CoV) of 7.3%.

CONCLUSION

Our results show promising results that a robust IDIF can be produced for dynamic ¹⁵O-water PET scans using only the dynamic PET scan images with no need for a corresponding MRI or complex analytical techniques and thereby making routine clinical use of quantitative CBF measurements with ¹⁵O-water feasible.

PET image enhancement using artificial intelligence for better characterization of epilepsy lesions

INTRODUCTION

[18F]fluorodeoxyglucose ([18F]FDG) brain PET is used clinically to detect small areas of decreased uptake associated with epileptogenic lesions, e.g., Focal Cortical Dysplasias (FCD) but its performance is limited due to spatial resolution and low contrast. We aimed to develop a deep learning-based PET image enhancement method using simulated PET to improve lesion visualization.

METHODS

We created 210 numerical brain phantoms (MRI segmented into 9 regions) and assigned 10 different plausible activity values (e.g., GM/WM ratios) resulting in 2100 ground truth high quality (GT-HQ) PET phantoms. With a validated Monte-Carlo PET simulator, we then created 2100 simulated standard quality (S-SQ) [18F]FDG scans. We trained a ResNet on 80% of this dataset (10% used for validation) to learn the mapping between S-SQ and GT-HQ PET, outputting a predicted HQ (P-HQ) PET. For the remaining 10%, we assessed Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), and Root Mean Squared Error (RMSE) against GT-HQ PET. For GM and WM, we computed recovery coefficients (RC) and coefficient of variation (COV). We also created lesioned GT-HQ phantoms, S-SQ PET and P-HQ PET with simulated small hypometabolic lesions characteristic of FCDs. We evaluated lesion detectability on S-SQ and P-HQ PET both visually and measuring the Relative Lesion Activity (RLA, measured activity in the reduced-activity ROI over the standard-activity ROI). Lastly, we applied our previously trained ResNet on 10 clinical epilepsy PETs to predict the corresponding HQ-PET and assessed image quality and confidence metrics.

RESULTS

Compared to S-SQ PET, P-HQ PET improved PNSR, SSIM and RMSE; significatively improved GM RCs (from 0.29 ± 0.03 to 0.79 ± 0.04) and WM RCs (from 0.49 ± 0.03 to 1 ± 0.05); mean COVs were not statistically different. Visual lesion detection improved from 38 to 75%, with average RLA decreasing from 0.83 ± 0.08 to 0.67 ± 0.14. Visual quality of P-HQ clinical PET improved as well as reader confidence.

CONCLUSION

P-HQ PET showed improved image quality compared to S-SQ PET across several objective quantitative metrics and increased detectability of simulated lesions. In addition, the model generalized to clinical data. Further evaluation is required to study generalization of our method and to assess clinical performance in larger cohorts.

Impact of spill-in counts from off-target regions on [18F]Flortaucipir PET quantification

BACKGROUND

[¹⁸F]Flortaucipir (FTP) PET quantification is usually hindered by spill-in counts from off-target binding (OFF) regions. The present work aims to provide an in-depth analysis of the impact of OFF in FTP PET quantification, as well as to identify optimal partial volume correction (PVC) strategies to minimize this problem.

METHODS

309 amyloid-beta (Aβ) negative cognitively normal subjects were included in the study. Additionally, 510 realistic FTP images with different levels of OFF were generated using Monte Carlo simulation (MC). Images were corrected for PVC using both a simple two-compartment and a multi-region method including OFF regions. FTP standardized uptake value ratio (SUVR) was quantified in Braak Areas (BA), the hippocampus (which was not included in Braak I/II) and different OFF regions (caudate, putamen, pallidum, thalamus, choroid plexus (ChPlex), cerebellar white matter (cerebWM), hemispheric white matter (hemisWM) and cerebrospinal fluid (CSF)) using the lower portion of the cerebellum as a reference region. The correlations between OFF and cortical SUVRs were studied both in real and in simulated PET images, with and without PVC.

RESULTS

In-vivo, we found correlations between all OFF and target regions, especially strong for the hemisWM (slope>0.63, R²>0.4). All the correlations were attenuated but remained significant after applying PVC, except for the ChPlex. In MC simulations, the hemisWM and CSF were the main contributors to PVE in all BA (slopes 0.15-0.26 and 0.13-0.21 respectively). The hemisWM (slope=0.2), as well as the ChPlex (slope=0.02), influenced SUVRs in the hippocampus. The CerebWM was negatively correlated with all target regions (slope<-0.02, R²>0.8). While no other correlations between OFF and target regions were found, hemisWM was correlated with all OFF regions but the cerebWM (slopes 0.06-0.33). HemisWM correlations attenuated (slopes<0.06) when applying two-compartment PVC, but the hippocampus-ChPlex and the cerebWM correlations required more complex PVC with dedicated compartments for these regions. In-vivo, PVC removed a notably higher fraction of the correlation between OFF regions found to be affected by PVE in the simulation studies and BA (≈50%) than for OFF regions not affected by PVE (16%).

CONCLUSION

HemisWM is the main driver of spill-in effects in FTP PET, affecting both target regions and the rest of OFF regions. PVC successfully reduces PVE, even when using a simple two-compartment method. Despite PVC, non-zero correlations were still observed between target and OFF regions in vivo, which suggests the existence of biological or tracer-related contributions to these correlations.

Relationship between astrocyte reactivity, using novel 11C-BU99008 PET, and glucose metabolism, grey matter volume and amyloid load in cognitively impaired individuals

Post mortem neuropathology suggests that astrocyte reactivity may play a significant role in neurodegeneration in Alzheimer's disease. We explored this in vivo using multimodal PET and MRI imaging. Twenty subjects (11 older, cognitively impaired patients and 9 age-matched healthy controls) underwent brain scanning using the novel reactive astrocyte PET tracer ¹¹C-BU99008, ¹⁸F-FDG and ¹⁸F-florbetaben PET, and T1-weighted MRI. Differences between cognitively impaired patients and healthy controls in regional and voxel-wise levels of astrocyte reactivity, glucose metabolism, grey matter volume and amyloid load were explored, and their relationship to each other was assessed using Biological Parametric Mapping (BPM). Amyloid beta (Aβ)-positive patients showed greater ¹¹C-BU99008 uptake compared to controls, except in the temporal lobe, whilst further increased ¹¹C-BU99008 uptake was observed in Mild Cognitive Impairment subjects compared to those with Alzheimer's disease in the frontal, temporal and cingulate cortices. BPM correlations revealed that regions which showed reduced ¹¹C-BU99008 uptake in Aβ-positive patients compared to controls, such as the temporal lobe, also showed reduced ¹⁸F-FDG uptake and grey matter volume, although the correlations with ¹⁸F-FDG uptake were not replicated in the ROI analysis. BPM analysis also revealed a regionally-dynamic relationship between astrocyte reactivity and amyloid uptake: increased amyloid load in cortical association areas of the temporal lobe and cingulate cortices was associated with reduced ¹¹C-BU99008 uptake, whilst increased amyloid uptake in primary motor and sensory areas (in which amyloid deposition occurs later) was associated with increased ¹¹C-BU99008 uptake. These novel observations add to the hypothesis that while astrocyte reactivity may be triggered by early Aβ-deposition, sustained pro-inflammatory astrocyte reactivity with greater amyloid deposition may lead to astrocyte dystrophy and amyloid-associated neuropathology such as grey matter atrophy and glucose hypometabolism, although the evidence for glucose hypometabolism here is less strong.

Astrocyte activation imaging with 11C-acetate and amyloid PET in mild cognitive impairment due to Alzheimer pathology

BACKGROUND

Neuroinflammation is a well-known feature of early Alzheimer disease (AD) yet astrocyte activation has not been extensively evaluated with in vivo imaging in mild cognitive impairment (MCI) due to amyloid plaque pathology. Unlike neurons, astrocytes metabolize acetate, which has potential as a glial biomarker in neurodegeneration in response to AD pathologic features. Since the medial temporal lobe (MTL) is a hotspot for AD neurodegeneration and inflammation, we assessed astrocyte activity in the MTL and compared it to amyloid and cognition.

METHODS

We evaluate spatial patterns of in vivo astrocyte activation and their relationships to amyloid deposition and cognition in a cross-sectional pilot study of six participants with MCI and five cognitively normal participants. We measure 11C-acetate and 18F-florbetaben amyloid standardized uptake values ratios (SUVRs) and kinetic flux compared to the cerebellum on PET, with MRI and neurocognitive testing.

RESULTS

MTL 11C-acetate SUVR was significantly elevated in MCI compared to cognitively normal participants (P = 0.03; Cohen d = 1.76). Moreover, MTL 11C-acetate SUVR displayed significant associations with global and regional amyloid burden in MCI. Greater MTL 11C-acetate retention was significantly related with worse neurocognitive measures including the Montreal Cognitive Assessment (P = 0.001), word list recall memory (P = 0.03), Boston naming test (P = 0.04) and trails B test (P = 0.04).

CONCLUSIONS

While further validation is required, this exploratory pilot study suggests a potential role for 11C-acetate PET as a neuroinflammatory biomarker in MCI and early AD to provide clinical and translational insights into astrocyte activation as a pathological response to amyloid.

Characterization of point-spread function specification error on Geometric Transfer Matrix partial volume correction in [11C]PiB amyloid imaging

Purpose

Partial-volume correction (PVC) using the Geometric Transfer Matrix (GTM) method is used in positron emission tomography (PET) to compensate for the effects of spatial resolution on quantitation. We evaluate the effect of misspecification of scanner point-spread function (PSF) on GTM results in amyloid imaging, including the effect on amyloid status classification (positive or negative).

Methods

Twenty-nine subjects with Pittsburgh Compound B ([¹¹C]PiB) PET and structural T1 MR imaging were analyzed. FreeSurfer 5.3 (FS) was used to parcellate MR images into regions-of-interest (ROIs) that were used to extract radioactivity concentration values from the PET images. GTM PVC was performed using our “standard” PSF parameterization [3D Gaussian, full-width at half-maximum (w) of approximately 5 mm]. Additional GTM PVC was performed with “incorrect” parameterizations, taken around the correct value. The result is a set of regional activity values for each of the GTM applications. For each case, activity values from various ROIs were combined and normalized to produce standardized uptake value ratios (SUVRs) for nine standard [¹¹C]PiB quantitation ROIs and a global region. GTM operating-point characteristics were determined from the slope of apparent SUVR versus w curves.

Results

Errors in specification of w on the order of 1 mm (3D) mainly produce only modest errors of up to a few percent. An exception was the anterior ventral striatum in which fractional errors of up to 0.29 per millimeter (3D) of error in w were observed.

Conclusion

While this study does not address all the issues regarding the quantitative strengths and weakness of GTM PVC, we find that with reasonable caution, the unavoidable inaccuracies associated with PSF specification do not preclude its use in amyloid quantitation.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40658-021-00403-5.

Cerebrovascular reactivity measurements using simultaneous 15O-water PET and ASL MRI: Impacts of arterial transit time, labeling efficiency, and hematocrit

Cerebrovascular reactivity (CVR) reflects the capacity of the brain to meet changing physiological demands and can predict the risk of cerebrovascular diseases. CVR can be obtained by measuring the change in cerebral blood flow (CBF) during a brain stress test where CBF is altered by a vasodilator such as acetazolamide. Although the gold standard to quantify CBF is PET imaging, the procedure is invasive and inaccessible to most patients. Arterial spin labeling (ASL) is a non-invasive and quantitative MRI method to measure CBF, and a consensus guideline has been published for the clinical application of ASL. Despite single post labeling delay (PLD) pseudo-continuous ASL (PCASL) being the recommended ASL technique for CBF quantification, it is sensitive to variations to the arterial transit time (ATT) and labeling efficiency induced by the vasodilator in CVR studies. Multi-PLD ASL controls for the changes in ATT, and velocity selective ASL is in theory insensitive to both ATT and labeling efficiency. Here we investigate CVR using simultaneous ¹⁵O-water PET and ASL MRI data from 19 healthy subjects. CVR and CBF measured by the ASL techniques were compared using PET as the reference technique. The impacts of blood T1 and labeling efficiency on ASL were assessed using individual measurements of hematocrit and flow velocity data of the carotid and vertebral arteries measured using phase-contrast MRI. We found that multi-PLD PCASL is the ASL technique most consistent with PET for CVR quantification (group mean CVR of the whole brain = 42 ± 19% and 40 ± 18% respectively). Single-PLD ASL underestimated the CVR of the whole brain significantly by 15 ± 10% compared with PET (p<0.01, paired t-test). Changes in ATT pre- and post-acetazolamide was the principal factor affecting ASL-based CVR quantification. Variations in labeling efficiency and blood T1 had negligible effects.

Voxel-wise partial volume correction method for accurate estimation of tissue sodium concentration in 23 Na-MRI at 7 T

Sodium is crucial for the maintenance of cell physiology, and its regulation of the sodium-potassium pump has implications for various neurological conditions. The distribution of sodium concentrations in tissue can be quantitatively evaluated by means of sodium MRI (²³ Na-MRI). Despite its usefulness in diagnosing particular disease conditions, tissue sodium concentration (TSC) estimated from ²³ Na-MRI can be strongly biased by partial volume effects (PVEs) that are induced by broad point spread functions (PSFs) as well as tissue fraction effects. In this work, we aimed to propose a robust voxel-wise partial volume correction (PVC) method for ²³ Na-MRI. The method is based on a linear regression (LR) approach to correct for tissue fraction effects, but it utilizes a 3D kernel combined with a modified least trimmed square (3D-mLTS) method in order to minimize regression-induced inherent smoothing effects. We acquired ²³ Na-MRI data with conventional Cartesian sampling at 7 T, and spill-over effects due to the PSF were considered prior to correcting for tissue fraction effects using 3D-mLTS. In the simulation, we found that the TSCs of gray matter (GM) and white matter (WM) were underestimated by 20% and 11% respectively without correcting tissue fraction effects, but the differences between ground truth and PVE-corrected data after the PVC using the 3D-mLTS method were only approximately 0.6% and 0.4% for GM and WM, respectively. The capability of the 3D-mLTS method was further demonstrated with in vivo ²³ Na-MRI data, showing significantly lower regression errors (ie root mean squared error) as compared with conventional LR methods (p < 0.001). The results of simulation and in vivo experiments revealed that 3D-mLTS is superior for determining under- or overestimated TSCs while preserving anatomical details. This suggests that the 3D-mLTS method is well suited for the accurate determination of TSC, especially in small focal lesions associated with pathological conditions.

Advances in multimodal data fusion in neuroimaging: Overview, challenges, and novel orientation

Multimodal fusion in neuroimaging combines data from multiple imaging modalities to overcome the fundamental limitations of individual modalities. Neuroimaging fusion can achieve higher temporal and spatial resolution, enhance contrast, correct imaging distortions, and bridge physiological and cognitive information. In this study, we analyzed over 450 references from PubMed, Google Scholar, IEEE, ScienceDirect, Web of Science, and various sources published from 1978 to 2020. We provide a review that encompasses (1) an overview of current challenges in multimodal fusion (2) the current medical applications of fusion for specific neurological diseases, (3) strengths and limitations of available imaging modalities, (4) fundamental fusion rules, (5) fusion quality assessment methods, and (6) the applications of fusion for atlas-based segmentation and quantification. Overall, multimodal fusion shows significant benefits in clinical diagnosis and neuroscience research. Widespread education and further research amongst engineers, researchers and clinicians will benefit the field of multimodal neuroimaging.

A Formula to Calculate the Threshold for Radiotherapy Targets on PET Images: Simulation Study

Background

Positron emission tomography (PET) images are being applied for defining radiotherapy targets. However, a recognized method for defining radiotherapy targets is lacking. We investigate the threshold to outline the radiotherapy target of a tumor on PET images and its influencing factors, and then expressed it by formula.

Methods

PET imaging for spherical tumors with a different tumor diameter (D), under different system resolutions [full width at half maximum (FWHM)], in different backgrounds with different pixel sizes, was simulated. PET images were analyzed to determine the relationship between the threshold and the factors mentioned above. Finally, the simulation results were verified by phantom experiments.

Results

The threshold decreased sharply with D for D < 2 FWHM, reached the minimum of 31% at D = 2 FWHM and then increased slowly, and it tended to constant for D > 8 FWHM. The threshold decreased with FWHM for FWHM < D/2, reached a minimum at FWHM = D/2, and then increased. The threshold increased with pixel size for D ≤ FWHM and decreased for D > FWHM. The threshold was independent of the background. The relationship between the threshold and its influencing factors was expressed as a formula. The results of the phantom verification indicated that the error of the target volume delineation that was calculated by the formula was less than 9%.

Conclusions

The threshold changes with tumor size, resolution of the PET system and pixel size according to certain rules. The formula to calculate the threshold could provide a method to estimate threshold to outline the radiotherapy target (tumor).

Predicting 15O-Water PET cerebral blood flow maps from multi-contrast MRI using a deep convolutional neural network with evaluation of training cohort bias

To improve the quality of MRI-based cerebral blood flow (CBF) measurements, a deep convolutional neural network (dCNN) was trained to combine single- and multi-delay arterial spin labeling (ASL) and structural images to predict gold-standard ¹⁵O-water PET CBF images obtained on a simultaneous PET/MRI scanner. The dCNN was trained and tested on 64 scans in 16 healthy controls (HC) and 16 cerebrovascular disease patients (PT) with 4-fold cross-validation. Fidelity to the PET CBF images and the effects of bias due to training on different cohorts were examined. The dCNN significantly improved CBF image quality compared with ASL alone (mean ± standard deviation): structural similarity index (0.854 ± 0.036 vs. 0.743 ± 0.045 [single-delay] and 0.732 ± 0.041 [multi-delay], P < 0.0001); normalized root mean squared error (0.209 ± 0.039 vs. 0.326 ± 0.050 [single-delay] and 0.344 ± 0.055 [multi-delay], P < 0.0001). The dCNN also yielded mean CBF with reduced estimation error in both HC and PT (P < 0.001), and demonstrated better correlation with PET. The dCNN trained with the mixed HC and PT cohort performed the best. The results also suggested that models should be trained on cases representative of the target population.

Emphasis Learning, Features Repetition in Width Instead of Length to Improve Classification Performance: Case Study-Alzheimer's Disease Diagnosis

In the past decade, many studies have been conducted to advance computer-aided systems for Alzheimer's disease (AD) diagnosis. Most of them have recently developed systems concentrated on extracting and combining features from MRI, PET, and CSF. For the most part, they have obtained very high performance. However, improving the performance of a classification problem is complicated, specifically when the model's accuracy or other performance measurements are higher than 90%. In this study, a novel methodology is proposed to address this problem, specifically in Alzheimer's disease diagnosis classification. This methodology is the first of its kind in the literature, based on the notion of replication on the feature space instead of the traditional sample space. Briefly, the main steps of the proposed method include extracting, embedding, and exploring the best subset of features. For feature extraction, we adopt VBM-SPM; for embedding features, a concatenation strategy is used on the features to ultimately create one feature vector for each subject. Principal component analysis is applied to extract new features, forming a low-dimensional compact space. A novel process is applied by replicating selected components, assessing the classification model, and repeating the replication until performance divergence or convergence. The proposed method aims to explore most significant features and highest-preforming model at the same time, to classify normal subjects from AD and mild cognitive impairment (MCI) patients. In each epoch, a small subset of candidate features is assessed by support vector machine (SVM) classifier. This repeating procedure is continued until the highest performance is achieved. Experimental results reveal the highest performance reported in the literature for this specific classification problem. We obtained a model with accuracies of 98.81%, 81.61%, and 81.40% for AD vs. normal control (NC), MCI vs. NC, and AD vs. MCI classification, respectively.

Spill-in counts in the quantification of 18F-florbetapir on Aβ-negative subjects: the effect of including white matter in the reference region

Background

We aim to provide a systematic study of the impact of white matter (WM) spill-in on the calculation of standardized uptake value ratios (SUVRs) on Aβ-negative subjects, and we study the effect of including WM in the reference region as a compensation. In addition, different partial volume correction (PVC) methods are applied and evaluated.

Methods

We evaluated magnetic resonance imaging and ¹⁸F-AV-45 positron emission tomography data from 122 cognitively normal (CN) patients recruited at the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Cortex SUVRs were obtained by using the cerebellar grey matter (CGM) (SUVR_CGM) and the whole cerebellum (SUVR_WC) as reference regions. The correlations between the different SUVRs and the WM uptake (WM-SUVR_CGM) were studied in patients, and in a well-controlled framework based on Monte Carlo (MC) simulation. Activity maps for the MC simulation were derived from ADNI patients by using a voxel-wise iterative process (BrainViset). Ten WM uptakes covering the spectrum of WM values obtained from patient data were simulated for different patients. Three different PVC methods were tested (a) the regional voxel-based (RBV), (b) the iterative Yang (iY), and (c) a simplified analytical correction derived from our MC simulation.

Results

WM-SUVR_CGM followed a normal distribution with an average of 1.79 and a standard deviation of 0.243 (13.6%). SUVR_CGM was linearly correlated to WM-SUVR_CGM (r = 0.82, linear fit slope = 0.28). SUVR_WC was linearly correlated to WM-SUVR_CGM (r = 0.64, linear fit slope = 0.13). Our MC results showed that these correlations are compatible with those produced by isolated spill-in effect (slopes of 0.23 and 0.11). The impact of the spill-in was mitigated by using PVC for SUVR_CGM (slopes of 0.06 and 0.07 for iY and RBV), while SUVR_WC showed a negative correlation with SUVR_CGM after PVC. The proposed analytical correction also reduced the observed correlations when applied to patient data (r = 0.27 for SUVR_CGM, r = 0.18 for SUVR_WC).

Conclusions

There is a high correlation between WM uptake and the measured SUVR due to spill-in effect, and that this effect is reduced when including WM in the reference region. We also evaluated the performance of PVC, and we proposed an analytical correction that can be applied to preprocessed data.

Spatially-variant image-based modeling of PSF deformations with application to a limited angle geometry from a dual-panel breast-PET imager

Dual-panel PET system configuration can lead to spatially variable point-spread functions (PSF) of considerable deformations due to depth-of-interaction effects and limited angular coverage. If not modelled properly, these effects result in decreased and inconsistent recovery of lesion activity across the field-of-view (FOV), as well as mispositioning of lesions in the reconstructed image caused by strong PSF asymmetries. We implemented and evaluated models of such PSF deformations with spatially-variant image-based resolution modeling (IRM) within reconstruction (varRM) using the Direct Image REConstruction for Time-of-flight (DIRECT) method and within post-reconstruction deconvolution methods. In addition, DIRECT reconstruction was performed with a spatially-invariant IRM (invRM) and without resolution modeling (noRM) for comparison. The methods were evaluated using simulated data for a realistic breast model with a set of 5 mm lesions located throughout the FOV of a dual-panel Breast-PET scanner. We simulated high-count data to focus on the ability of each method to correctly recover the PSF deformations, and a clinically realistic count level to assess the impact of low count data on the quantitative performance of the evaluated techniques. Performance of the methods evaluated herein was assessed by comparing lesion activity recovery (%BIAS), consistency (%SD) across the FOV, overall error (%RMSE), and recovery of each lesion location. As expected, all techniques using IRM provide considerable improvement over the noRM reconstruction. For the high-count cases, the overall quantitative performance of all IRM techniques, whether within reconstruction or within post-reconstruction, is similar if the lesion location misplacements are ignored. However, invRM provides less consistent performance on activity across lesions and is not able to recover accurate lesion locations. For a clinically realistic count level, varRM reconstruction consistently outperforms all compared approaches, while the post-reconstruction IRM approaches exhibit higher %SD and %RMSE values due to being more affected by the data noise than the within-reconstruction IRM approaches.

Noninvasive quantification of macrophagic lung recruitment during experimental ventilation-induced lung injury

Macrophagic lung infiltration is pivotal in the development of lung biotrauma because of ventilation-induced lung injury (VILI). We assessed the performance of [¹¹C](R)-PK11195, a positron emission tomography (PET) radiotracer binding the translocator protein, to quantify macrophage lung recruitment during experimental VILI. Pigs (n = 6) were mechanically ventilated under general anesthesia, using protective ventilation settings (baseline). Experimental VILI was performed by titrating tidal volume to reach a transpulmonary end-inspiratory pressure (∆P_L) of 35-40 cmH₂O. We acquired PET/computed tomography (CT) lung images at baseline and after 4 h of VILI. Lung macrophages were quantified in vivo by the standardized uptake value (SUV) of [¹¹C](R)-PK11195 measured in PET on the whole lung and in six lung regions and ex vivo on lung pathology at the end of experiment. Lung mechanics were extracted from CT images to assess their association with the PET signal. ∆P_L increased from 9 ± 1 cmH₂O under protective ventilation, to 36 ± 6 cmH₂O during experimental VILI. Compared with baseline, whole-lung [¹¹C](R)-PK11195 SUV significantly increased from 1.8 ± 0.5 to 2.9 ± 0.5 after experimental VILI. Regional [¹¹C](R)-PK11195 SUV was positively associated with the magnitude of macrophage recruitment in pathology (P = 0.03). Compared with baseline, whole-lung CT-derived dynamic strain and tidal hyperinflation increased significantly after experimental VILI, from 0.6 ± 0 to 2.0 ± 0.4, and 1 ± 1 to 43 ± 19%, respectively. On multivariate analysis, both were significantly associated with regional [¹¹C](R)-PK11195 SUV. [¹¹C](R)-PK11195 lung uptake (a proxy of lung inflammation) was increased by experimental VILI and was associated with the magnitude of dynamic strain and tidal hyperinflation.NEW & NOTEWORTHY We assessed the performance of [¹¹C](R)-PK11195, a translocator protein-specific positron emission tomography (PET) radiotracer, to quantify macrophage lung recruitment during experimental ventilation-induced lung injury (VILI). In this proof-of-concept study, we showed that the in vivo quantification of [¹¹C](R)-PK11195 lung uptake in PET reflected the magnitude of macrophage lung recruitment after VILI. Furthermore, increased [¹¹C](R)-PK11195 lung uptake was associated with harmful levels of dynamic strain and tidal hyperinflation applied to the lungs.

Evaluating the In Vivo Specificity of [18F]UCB-H for the SV2A Protein, Compared with SV2B and SV2C in Rats Using microPET

The synaptic vesicle protein 2 (SV2) is involved in synaptic vesicle trafficking. The SV2A isoform is the most studied and its implication in epilepsy therapy led to the development of the first SV2A PET radiotracer [¹⁸F]UCB-H. The objective of this study was to evaluate in vivo, using microPET in rats, the specificity of [¹⁸F]UCB-H for SV2 isoform A in comparison with the other two isoforms (B and C) through a blocking assay. Twenty Sprague Dawley rats were pre-treated either with the vehicle, or with specific competitors against SV2A (levetiracetam), SV2B (UCB5203) and SV2C (UCB0949). The distribution volume (Vt, Logan plot, t* 15 min) was obtained with a population-based input function. The Vt analysis for the entire brain showed statistically significant differences between the levetiracetam group and the other groups (p < 0.001), but also between the vehicle and the SV2B group (p < 0.05). An in-depth Vt analysis conducted for eight relevant brain structures confirmed the statistically significant differences between the levetiracetam group and the other groups (p < 0.001) and highlighted the superior and the inferior colliculi along with the cortex as regions also displaying statistically significant differences between the vehicle and SV2B groups (p < 0.05). These results emphasize the in vivo specificity of [¹⁸F]UCB-H for SV2A against SV2B and SV2C, confirming that [¹⁸F]UCB-H is a suitable radiotracer for in vivo imaging of the SV2A proteins with PET.

Effects of systematic partial volume errors on the estimation of gray matter cerebral blood flow with arterial spin labeling MRI

OBJECTIVE

Partial volume (PV) correction is an important step in arterial spin labeling (ASL) MRI that is used to separate perfusion from structural effects when computing the mean gray matter (GM) perfusion. There are three main methods for performing this correction: (1) GM-threshold, which includes only voxels with GM volume above a preset threshold; (2) GM-weighted, which uses voxel-wise GM contribution combined with thresholding; and (3) PVC, which applies a spatial linear regression algorithm to estimate the flow contribution of each tissue at a given voxel. In all cases, GM volume is obtained using PV maps extracted from the segmentation of the T1-weighted (T1w) image. As such, PV maps contain errors due to the difference in readout type and spatial resolution between ASL and T1w images. Here, we estimated these errors and evaluated their effect on the performance of each PV correction method in computing GM cerebral blood flow (CBF).

MATERIALS AND METHODS

Twenty-two volunteers underwent scanning using 2D echo planar imaging (EPI) and 3D spiral ASL. For each PV correction method, GM CBF was computed using PV maps simulated to contain estimated errors due to spatial resolution mismatch and geometric distortions which are caused by the mismatch in readout between ASL and T1w images. Results were analyzed to assess the effect of each error on the estimation of GM CBF from ASL data.

RESULTS

Geometric distortion had the largest effect on the 2D EPI data, whereas the 3D spiral was most affected by the resolution mismatch. The PVC method outperformed the GM-threshold even in the presence of combined errors from resolution mismatch and geometric distortions. The quantitative advantage of PVC was 16% without and 10% with the combined errors for both 2D and 3D ASL. Consistent with theoretical expectations, for error-free PV maps, the PVC method extracted the true GM CBF. In contrast, GM-weighted overestimated GM CBF by 5%, while GM-threshold underestimated it by 16%. The presence of PV map errors decreased the calculated GM CBF for all methods.

CONCLUSION

The quality of PV maps presents no argument for the preferential use of the GM-threshold method over PVC in the clinical application of ASL.

Increased Striatal Dopamine Synthesis Capacity in Gambling Addiction

BACKGROUND

The hypothesis that dopamine plays an important role in the pathophysiology of pathological gambling is pervasive. However, there is little to no direct evidence for a categorical difference between pathological gamblers and healthy control subjects in terms of dopamine transmission in a drug-free state. Here we provide evidence for this hypothesis by comparing dopamine synthesis capacity in the dorsal and ventral parts of the striatum in 13 pathological gamblers and 15 healthy control subjects.

METHODS

This was achieved using [¹⁸F]fluoro-levo-dihydroxyphenylalanine dynamic positron emission tomography scans and striatal regions of interest that were hand-drawn based on visual inspection of individual structural magnetic resonance imaging scans.

RESULTS

Our results show that dopamine synthesis capacity was increased in pathological gamblers compared with healthy control subjects. Dopamine synthesis was 16% higher in the caudate body, 17% higher in the dorsal putamen, and 17% higher in the ventral striatum in pathological gamblers compared with control subjects. Moreover, dopamine synthesis capacity in the dorsal putamen and caudate head was positively correlated with gambling distortions in pathological gamblers.

CONCLUSIONS

Taken together, these results provide empirical evidence for increased striatal dopamine synthesis in pathological gambling.

Quantification of FDG-PET/CT with delayed imaging in patients with newly diagnosed recurrent breast cancer

Background

Several studies have shown the advantage of delayed-time-point imaging with 18F-FDG-PET/CT to distinguish malignant from benign uptake. This may be relevant in cancer diseases with low metabolism, such as breast cancer. We aimed at examining the change in SUV from 1 h (1h) to 3 h (3h) time-point imaging in local and distant lesions in patients with recurrent breast cancer. Furthermore, we investigated the effect of partial volume correction in the different types of metastases, using semi-automatic quantitative software (ROVER™).

Methods

One-hundred and two patients with suspected breast cancer recurrence underwent whole-body PET/CT scans 1h and 3h after FDG injection. Semi-quantitative standardised uptake values (SUVmax, SUVmean) and partial volume corrected SUVmean (cSUVmean), were estimated in malignant lesions, and as reference in healthy liver tissue. The change in quantitative measures from 1h to 3h was calculated, and SUVmean was compared to cSUVmean. Metastases were verified by biopsy.

Results

Of the 102 included patients, 41 had verified recurrent disease with in median 15 lesions (range 1-70) amounting to a total of 337 malignant lesions included in the analysis. SUVmax of malignant lesions increased from 6.4 ± 3.4 [0.9-19.7] (mean ± SD, min and max) at 1h to 8.1 ± 4.4 [0.7-29.7] at 3h. SUVmax in breast, lung, lymph node and bone lesions increased significantly (p < 0.0001) between 1h and 3h by on average 25, 40, 33, and 27%, respectively. A similar pattern was observed with (uncorrected) SUVmean. Partial volume correction increased SUVmean significantly, by 63 and 71% at 1h and 3h imaging, respectively. The highest impact was in breast lesions at 3h, where cSUVmean increased by 87% compared to SUVmean.

Conclusion

SUVs increased from 1h to 3h in malignant lesions, SUVs of distant recurrence were in general about twice as high as those of local recurrence. Partial volume correction caused significant increases in these values. However, it is questionable, if these relatively modest quantitative advances of 3h imaging are sufficient to warrant delayed imaging in this patient group.

Trial registration

ClinicalTrails.gov NCT01552655. Registered 28 February 2012, partly retrospectively registered.

Electronic supplementary material

The online version of this article (10.1186/s12880-018-0254-8) contains supplementary material, which is available to authorized users.

Multimodal partial volume correction: Application to [11C]PIB PET/MRI myelin imaging in multiple sclerosis

The [11C]PIB PET tracer, originally developed for amyloid imaging, has been recently repurposed to quantify demyelination and remyelination in multiple sclerosis (MS). Myelin PET imaging, however, is limited by its low resolution that deteriorates the quantification accuracy of white matter (WM) lesions. Here, we introduce a novel partial volume correction (PVC) method called Multiresolution-Multimodal Resolution-Recovery (MM-RR), which uses the wavelet transform and a synergistic statistical model to exploit MRI structural images to improve the resolution of [11C]PIB PET myelin imaging. MM-RR performance was tested on a phantom acquisition and in a dataset comprising [11C]PIB PET and MR T1- and T2-weighted images of 8 healthy controls and 20 MS patients. For the control group, the MM-RR PET images showed an average increase of 5.7% in WM uptake while the grey-matter (GM) uptake remained constant, resulting in +31% WM/GM contrast. Furthermore, MM-RR PET binding maps correlated significantly with the mRNA expressions of the most represented proteins in the myelin sheath (R2 = 0.57 ± 0.09). In the patient group, MM-RR PET images showed sharper lesion contours and significant improvement in normal-appearing tissue/WM-lesion contrast compared to standard PET (contrast improvement > +40%). These results were consistent with MM-RR performances in phantom experiments.

Partial volume correction for PET quantification and its impact on brain network in Alzheimer's disease

Amyloid positron emission tomography (PET) imaging is a valuable tool for research and diagnosis in Alzheimer’s disease (AD). Partial volume effects caused by the limited spatial resolution of PET scanners degrades the quantitative accuracy of PET image. In this study, we have applied a method to evaluate the impact of a joint-entropy based partial volume correction (PVC) technique on brain networks learned from a clinical dataset of AV-45 PET image and compare network properties of both uncorrected and corrected image-based brain networks. We also analyzed the region-wise SUVRs of both uncorrected and corrected images. We further performed classification tests on different groups using the same set of algorithms with same parameter settings. PVC has sometimes been avoided due to increased noise sensitivity in image registration and segmentation, however, our results indicate that appropriate PVC may enhance the brain network structure analysis for AD progression and improve classification performance.

Differential Associations between Cortical Thickness and Striatal Dopamine in Treatment-Naïve Adults with ADHD vs. Healthy Controls

Alterations in catecholamine signaling and cortical morphology have both been implicated in the pathophysiology of attention deficit/hyperactivity disorder (ADHD). However, possible links between the two remain unstudied. Here, we report exploratory analyses of cortical thickness and its relation to striatal dopamine transmission in treatment-naïve adults with ADHD and matched healthy controls. All participants had one magnetic resonance imaging (MRI) and two [¹¹C]raclopride positron emission tomography scans. Associations between frontal cortical thickness and the magnitude of d-amphetamine-induced [¹¹C]raclopride binding changes were observed that were divergent in the two groups. In the healthy controls, a thicker cortex was associated with less dopamine release; in the ADHD participants the converse was seen. The same divergence was seen for baseline D2/3 receptor availability. In healthy volunteers, lower D2/3 receptor availability was associated with a thicker cortex, while in the ADHD group lower baseline D2/3 receptor availability was associated with a thinner cortex. Individual differences in cortical thickness in these regions correlated with ADHD symptom severity. Together, these findings add to the evidence of associations between dopamine transmission and cortical morphology, and suggest that these relationships are altered in treatment-naïve adults with ADHD.

Two weeks of moderate-intensity continuous training, but not high-intensity interval training, increases insulin-stimulated intestinal glucose uptake

This is the first study where the effects of exercise training on the intestinal substrate uptake have been investigated using the most advanced techniques available. We also show the importance of exercise intensity in inducing these changes.

Similar to muscles, the intestine is also insulin resistant in obese subjects and subjects with impaired glucose tolerance. Exercise training improves muscle insulin sensitivity, but its effects on intestinal metabolism are not known. We studied the effects of high-intensity interval training (HIIT) and moderate-intensity continuous training (MICT) on intestinal glucose and free fatty acid uptake from circulation in humans. Twenty-eight healthy, middle-aged, sedentary men were randomized for 2 wk of HIIT or MICT. Intestinal insulin-stimulated glucose uptake and fasting free fatty acid uptake from circulation were measured using positron emission tomography and [¹⁸F]FDG and [¹⁸F]FTHA. In addition, effects of HIIT and MICT on intestinal GLUT2 and CD36 protein expression were studied in rats. Training improved aerobic capacity (P = 0.001) and whole body insulin sensitivity (P = 0.04), but not differently between HIIT and MICT. Insulin-stimulated glucose uptake increased only after the MICT in the colon (HIIT = 0%; MICT = 37%) (P = 0.02 for time × training) and tended to increase in the jejunum (HIIT = −4%; MICT = 13%) (P = 0.08 for time × training). Fasting free fatty acid uptake decreased in the duodenum in both groups (HIIT = −6%; MICT = −48%) (P = 0.001 time) and tended to decrease in the colon in the MICT group (HIIT = 0%; MICT = −38%) (P = 0.08 for time × training). In rats, both training groups had higher GLUT2 and CD36 expression compared with control animals. This study shows that already 2 wk of MICT enhances insulin-stimulated glucose uptake, while both training modes reduce fasting free fatty acid uptake in the intestine in healthy, middle-aged men, providing an additional mechanism by which exercise training can improve whole body metabolism.

NEW & NOTEWORTHY This is the first study where the effects of exercise training on the intestinal substrate uptake have been investigated using the most advanced techniques available. We also show the importance of exercise intensity in inducing these changes.

Resolution Improvement in Positron Emission Tomography Using Anatomical Magnetic Resonance Imaging

An ideal imaging system should provide information with high-sensitivity, high spatial, and temporal resolution. Unfortunately, it is not possible to satisfy all of these desired features in a single modality. In this paper, we discuss methods to improve the spatial resolution in positron emission imaging (PET) using a priori information from Magnetic Resonance Imaging (MRI). Our approach uses an image restoration algorithm based on the maximization of mutual information (MMI), which has found significant success for optimizing multimodal image registration. The MMI criterion is used to estimate the parameters in the Sharpness-Constrained Wiener filter. The generated filter is then applied to restore PET images of a realistic digital brain phantom. The resulting restored images show improved resolution and better signal-to-noise ratio compared to the interpolated PET images. We conclude that a Sharpness-Constrained Wiener filter having parameters optimized from a MMI criterion may be useful for restoring spatial resolution in PET based on a priori information from correlated MRI.

Ensembles of Deep Learning Architectures for the Early Diagnosis of the Alzheimer's Disease

Computer Aided Diagnosis (CAD) constitutes an important tool for the early diagnosis of Alzheimer's Disease (AD), which, in turn, allows the application of treatments that can be simpler and more likely to be effective. This paper explores the construction of classification methods based on deep learning architectures applied on brain regions defined by the Automated Anatomical Labeling (AAL). Gray Matter (GM) images from each brain area have been split into 3D patches according to the regions defined by the AAL atlas and these patches are used to train different deep belief networks. An ensemble of deep belief networks is then composed where the final prediction is determined by a voting scheme. Two deep learning based structures and four different voting schemes are implemented and compared, giving as a result a potent classification architecture where discriminative features are computed in an unsupervised fashion. The resulting method has been evaluated using a large dataset from the Alzheimer's disease Neuroimaging Initiative (ADNI). Classification results assessed by cross-validation prove that the proposed method is not only valid for differentiate between controls (NC) and AD images, but it also provides good performances when tested for the more challenging case of classifying Mild Cognitive Impairment (MCI) Subjects. In particular, the classification architecture provides accuracy values up to 0.90 and AUC of 0.95 for NC/AD classification, 0.84 and AUC of 0.91 for stable MCI/AD classification and 0.83 and AUC of 0.95 for NC/MCI converters classification.

Multimodal Partial-Volume Correction: Application to 18F-Fluoride PET/CT Bone Metastases Studies

(18)F-fluoride PET/CT offers the opportunity for accurate skeletal metastasis staging, compared with conventional imaging methods. (18)F-fluoride is a bone-specific tracer whose uptake depends on osteoblastic activity. Because of the resulting increase in bone mineralization and sclerosis, the osteoblastic process can also be detected morphologically in CT images. Although CT is characterized by high resolution, the potential of PET is limited by its lower spatial resolution and the resulting partial-volume effect. In this context, the synergy between PET and CT presents an opportunity to resolve this limitation using a novel multimodal approach called synergistic functional-structural resolution recovery (SFS-RR). Its performance is benchmarked against current resolution recovery technology using the point-spread function (PSF) of the scanner in the reconstruction procedure.

METHODS

The SFS-RR technique takes advantage of the multiresolution property of the wavelet transform applied to both functional and structural images to create a high-resolution PET image that exploits the structural information of CT. Although the method was originally conceived for PET/MR imaging of brain data, an ad hoc version for whole-body PET/CT is proposed here. Three phantom experiments and 2 datasets of metastatic bone (18)F-fluoride PET/CT images from primary prostate and breast cancer were used to test the algorithm performances. The SFS-RR images were compared with the manufacturer's PSF-based reconstruction using the standardized uptake value (SUV) and the metabolic volume as metrics for quantification.

RESULTS

When compared with standard PET images, the phantom experiments showed a bias reduction of 14% in activity and 1.3 cm(3) in volume estimates for PSF images and up to 20% and 2.5 cm(3) for the SFS-RR images. The SFS-RR images were characterized by a higher recovery coefficient (up to 60%) whereas noise levels remained comparable to those of standard PET. The clinical data showed an increase in the SUV estimates for SFS-RR images up to 34% for peak SUV and 50% for maximum SUV and mean SUV. Images were also characterized by sharper lesion contours and better lesion detectability.

CONCLUSION

The proposed methodology generates PET images with improved quantitative and qualitative properties. Compared with standard methods, SFS-RR provides superior lesion segmentation and quantification, which may result in more accurate tumor characterization.

Multimodal Discrimination of Alzheimer's Disease Based on Regional Cortical Atrophy and Hypometabolism

Structural MR image (MRI) and ¹⁸F-Fluorodeoxyglucose-positron emission tomography (FDG-PET) have been widely employed in diagnosis of both Alzheimer’s disease (AD) and mild cognitive impairment (MCI) pathology, which has led to the development of methods to distinguish AD and MCI from normal controls (NC). Synaptic dysfunction leads to a reduction in the rate of metabolism of glucose in the brain and is thought to represent AD progression. FDG-PET has the unique ability to estimate glucose metabolism, providing information on the distribution of hypometabolism. In addition, patients with AD exhibit significant neuronal loss in cerebral regions, and previous AD research has shown that structural MRI can be used to sensitively measure cortical atrophy. In this paper, we introduced a new method to discriminate AD from NC based on complementary information obtained by FDG and MRI. For accurate classification, surface-based features were employed and 12 predefined regions were selected from previous studies based on both MRI and FDG-PET. Partial least square linear discriminant analysis was employed for making diagnoses. We obtained 93.6% classification accuracy, 90.1% sensitivity, and 96.5% specificity in discriminating AD from NC. The classification scheme had an accuracy of 76.5% and sensitivity and specificity of 46.5% and 89.6%, respectively, for discriminating MCI from AD. Our method exhibited a superior classification performance compared with single modal approaches and yielded parallel accuracy to previous multimodal classification studies using MRI and FDG-PET.

Proliferative and Glycolytic Assessment of the Whole-Body Bone Marrow Compartment

OBJECTIVE

Quantitative assessment of active bone marrow (BM) in vivo is yet to be well-defined. This study aims to compare total body BM volume estimations obtained from use of both18F-FLT PET/CT and 18F-FDG PET/CT in order to consolidate higher cellular proliferation rates with imaging the highly active red BM in pancreatic cancer.

METHODS

This phase I pilot study includes seven patients with pancreatic cancers who underwent both 18F-FLT and 18F-FDG imaging each acquired within a week's duration. A CT-based classifier is used for segmenting bone into cortical and trabecular regions. The total BM volume is determined through statistical thresholding on PET activity found within the trabecular bone.

RESULTS

Results showed that 18F-FLT measures of red BM volume (RBV) were higher than those obtained from 18F-FDG (∆=89.21 ml). RBV obtained using 18F-FLT in males were found to have high correlation with measured weight (R2=0.61) and BMI (R2=0.70). The red BM fraction obtained from 18F-FLT was significantly different between males and females, with females showing much higher red bone matter within the trabecular bone (p<0.05). In contrast to 18F-FLT, 18F-FDG BM measurements showed that RBV was significantly different between males and females (p<0.05). Results also show that spinal activity SUV threshold for red BM segmentation is significantly different between 18F-FLT PET and 18F-FDG PET (p<0.05).

CONCLUSION

By combining 18F-FLT-PET and 18F-FDG-PET, this study provides useful insights for in vivo BM estimation through its proliferative and glycolytic activities.

Simulation-based partial volume correction for dopaminergic PET imaging: Impact of segmentation accuracy

AIM

Partial volume correction (PVC) is an essential step for quantitative positron emission tomography (PET). In the present study, PVELab, a freely available software, is evaluated for PVC in (18)F-FDOPA brain-PET, with a special focus on the accuracy degradation introduced by various MR-based segmentation approaches.

METHODS

Four PVC algorithms (M-PVC; MG-PVC; mMG-PVC; and R-PVC) were analyzed on simulated (18)F-FDOPA brain-PET images. MR image segmentation was carried out using FSL (FMRIB Software Library) and SPM (Statistical Parametric Mapping) packages, including additional adaptation for subcortical regions (SPML). Different PVC and segmentation combinations were compared with respect to deviations in regional activity values and time-activity curves (TACs) of the occipital cortex (OCC), caudate nucleus (CN), and putamen (PUT). Additionally, the PVC impact on the determination of the influx constant (Ki) was assessed.

RESULTS

Main differences between tissue-maps returned by three segmentation algorithms were found in the subcortical region, especially at PUT. Average misclassification errors in combination with volume reduction was found to be lowest for SPML (PUT < 30%) and highest for FSL (PUT > 70%). Accurate recovery of activity data at OCC is achieved by M-PVC (apparent recovery coefficient varies between 0.99 and 1.10). The other three evaluated PVC algorithms have demonstrated to be more suitable for subcortical regions with MG-PVC and mMG-PVC being less prone to the largest tissue misclassification error simulated in this study. Except for M-PVC, quantification accuracy of Ki for CN and PUT was clearly improved by PVC.

CONCLUSIONS

The regional activity value of PUT was appreciably overcorrected by most of the PVC approaches employing FSL or SPM segmentation, revealing the importance of accurate MR image segmentation for the presented PVC framework. The selection of a PVC approach should be adapted to the anatomical structure of interest. Caution is recommended in subsequent interpretation of Ki values. The possible different change of activity concentrations due to PVC in both target and reference regions tends to alter the corresponding TACs, introducing bias to Ki determination. The accuracy of quantitative analysis was improved by PVC but at the expense of precision reduction, indicating the potential impropriety of applying the presented framework for group comparison studies.

Surface-based partial-volume correction for high-resolution PET

Tissue radioactivity concentrations, measured with positron emission tomography (PET) are subject to partial volume effects (PVE) due to the limited spatial resolution of the scanner. Last generation high-resolution PET cameras with a full width at half maximum (FWHM) of 2-4mm are less prone to PVEs than previous generations. Corrections for PVEs are still necessary, especially when studying small brain stem nuclei or small variations in cortical neuroreceptor concentrations which may be related to cytoarchitectonic differences. Although several partial-volume correction (PVC) algorithms exist, these are frequently based on a priori assumptions about tracer distribution or only yield corrected values of regional activity concentrations without providing PVE corrected images. We developed a new iterative deconvolution algorithm (idSURF) for PVC of PET images that aims to overcome these limitations by using two innovative techniques: 1) the incorporation of anatomic information from a cortical gray matter surface representation, extracted from magnetic resonance imaging (MRI) and 2) the use of anatomically constrained filtering to attenuate noise. PVE corrected images were generated with idSURF implemented into a non-interactive processing pipeline. idSURF was validated using simulated and clinical PET data sets and compared to a frequently used standard PVC method (Geometric Transfer Matrix: GTM). The results on simulated data sets show that idSURF consistently recovers accurate radiotracer concentrations within 1-5% of true values. Both radiotracer concentrations and non-displaceable binding potential (BPnd) values derived from clinical PET data sets with idSURF were highly correlated with those obtained with the standard PVC method (R(2) = 0.99, error = 0%-3.2%). These results suggest that idSURF is a valid and potentially clinically useful PVC method for automatic processing of large numbers of PET data sets.

Subclass-based multi-task learning for Alzheimer's disease diagnosis

In this work, we propose a novel subclass-based multi-task learning method for feature selection in computer-aided Alzheimer's Disease (AD) or Mild Cognitive Impairment (MCI) diagnosis. Unlike the previous methods that often assumed a unimodal data distribution, we take into account the underlying multipeak distribution of classes. The rationale for our approach is that it is highly likely for neuroimaging data to have multiple peaks or modes in distribution, e.g., mixture of Gaussians, due to the inter-subject variability. In this regard, we use a clustering method to discover the multipeak distributional characteristics and define subclasses based on the clustering results, in which each cluster covers a peak in the underlying multipeak distribution. Specifically, after performing clustering for each class, we encode the respective subclasses, i.e., clusters, with their unique codes. In encoding, we impose the subclasses of the same original class close to each other and those of different original classes distinct from each other. By setting the codes as new label vectors of our training samples, we formulate a multi-task learning problem in a ℓ2,1-penalized regression framework, through which we finally select features for classification. In our experimental results on the ADNI dataset, we validated the effectiveness of the proposed method by improving the classification accuracies by 1% (AD vs. Normal Control: NC), 3.25% (MCI vs. NC), 5.34% (AD vs. MCI), and 7.4% (MCI Converter: MCI-C vs. MCI Non-Converter: MCI-NC) compared to the competing single-task learning method. It is remarkable for the performance improvement in MCI-C vs. MCI-NC classification, which is the most important for early diagnosis and treatment. It is also noteworthy that with the strategy of modality-adaptive weights by means of a multi-kernel support vector machine, we maximally achieved the classification accuracies of 96.18% (AD vs. NC), 81.45% (MCI vs. NC), 73.21% (AD vs. MCI), and 74.04% (MCI-C vs. MCI-NC), respectively.

Amphetamine-induced dopamine release and neurocognitive function in treatment-naive adults with ADHD

Converging evidence from clinical, preclinical, neuroimaging, and genetic research implicates dopamine neurotransmission in the pathophysiology of attention deficit hyperactivity disorder (ADHD). The in vivo neuroreceptor imaging evidence also suggests alterations in the dopamine system in ADHD; however, the nature and behavioral significance of those have not yet been established. Here, we investigated striatal dopaminergic function in ADHD using [(11)C]raclopride PET with a d-amphetamine challenge. We also examined the relationship of striatal dopamine responses to ADHD symptoms and neurocognitive function. A total of 15 treatment-free, noncomorbid adult males with ADHD (age: 29.87 ± 8.65) and 18 healthy male controls (age: 25.44 ± 6.77) underwent two PET scans: one following a lactose placebo and the other following d-amphetamine (0.3 mg/kg, p.o.), administered double blind and in random order counterbalanced across groups. In a separate session without a drug, participants performed a battery of neurocognitive tests. Relative to the healthy controls, the ADHD patients, as a group, showed greater d-amphetamine-induced decreases in striatal [(11)C]raclopride binding and performed more poorly on measures of response inhibition. Across groups, a greater magnitude of d-amphetamine-induced change in [(11)C]raclopride binding potential was associated with poorer performance on measures of response inhibition and ADHD symptoms. Our findings suggest an augmented striatal dopaminergic response in treatment-naive ADHD. Though in contrast to results of a previous study, this finding appears consistent with a model proposing exaggerated phasic dopamine release in ADHD. A susceptibility to increased phasic dopamine responsivity may contribute to such characteristics of ADHD as poor inhibition and impulsivity.

Direct (17)O MRI with partial volume correction: first experiences in a glioblastoma patient

OBJECT

In tumor cells the energy production is shifted from aerobic to anaerobic metabolization of glucose, which makes the cerebral metabolic rate of oxygen consumption (CMRO2) a diagnostic parameter for tissue viability. Direct oxygen-17 ((17)O) MRI during inhalation of (17)O gas allows for a non-invasive determination of the CMRO2. However, the low spatial resolution and the fast transverse relaxation of (17)O lead to partial volume effects that severely bias the quantification of signal intensities. The aim of this work was to determine the CMRO2 in a tumor patient by (17)O MRI in combination with a partial volume correction (PVC) scheme.

MATERIALS AND METHODS

Direct (17)O MRI was performed in a glioblastoma patient (F, 51 years) prior to surgery at 7 T. The 'geometric transfer matrix' algorithm for volume of interest based PVC was adapted to (17)O MRI to recover the true signal intensities. We determined the CMRO2 values of gray matter (GM), white matter (WM), cerebrospinal fluid (CSF) and the tumor areas of the contrast enhancing rim (CE), the necrotic center (NE), and the perifocal edema (PE) using a three-phase metabolic model.

RESULTS

Large differences in the signal increase during (17)O2 inhalation were obtained ranging from less than 2% in the tumor center up to more than 20% in GM areas. After PVC of the signal time curves, we determined CMRO2 values of 0.67 ± 0.08 μmol/g/min (WM), 3.57 ± 0.67 μmol/g/min (GM), 0.35 ± 0.09 μmol/g/min (CE), and 0.42 ± 0.05 μmol/g/min (PE). In CSF and NE no oxygen uptake (i.e. CMRO2 = 0) was determined from the corrected signals, well in accordance with the underlying physiology in these regions.

CONCLUSION

The results show that PVC has a strong effect on the resulting CMRO2 values obtained by (17)O MRI. We found substantial differences-especially in GM tissue-between corrected and non-corrected CMRO2 values. Additionally, we demonstrated the feasibility of CMRO2 assessment in a glioblastoma patient by (17)O MRI.

Latent feature representation with stacked auto-encoder for AD/MCI diagnosis

Recently, there have been great interests for computer-aided diagnosis of Alzheimer's disease (AD) and its prodromal stage, mild cognitive impairment (MCI). Unlike the previous methods that considered simple low-level features such as gray matter tissue volumes from MRI, and mean signal intensities from PET, in this paper, we propose a deep learning-based latent feature representation with a stacked auto-encoder (SAE). We believe that there exist latent non-linear complicated patterns inherent in the low-level features such as relations among features. Combining the latent information with the original features helps build a robust model in AD/MCI classification, with high diagnostic accuracy. Furthermore, thanks to the unsupervised characteristic of the pre-training in deep learning, we can benefit from the target-unrelated samples to initialize parameters of SAE, thus finding optimal parameters in fine-tuning with the target-related samples, and further enhancing the classification performances across four binary classification problems: AD vs. healthy normal control (HC), MCI vs. HC, AD vs. MCI, and MCI converter (MCI-C) vs. MCI non-converter (MCI-NC). In our experiments on ADNI dataset, we validated the effectiveness of the proposed method, showing the accuracies of 98.8, 90.7, 83.7, and 83.3 % for AD/HC, MCI/HC, AD/MCI, and MCI-C/MCI-NC classification, respectively. We believe that deep learning can shed new light on the neuroimaging data analysis, and our work presented the applicability of this method to brain disease diagnosis.

Noise propagation in resolution modeled PET imaging and its impact on detectability

Positron emission tomography imaging is affected by a number of resolution degrading phenomena, including positron range, photon non-collinearity and inter-crystal blurring. An approach to this issue is to model some or all of these effects within the image reconstruction task, referred to as resolution modeling (RM). This approach is commonly observed to yield images of higher resolution and subsequently contrast, and can be thought of as improving the modulation transfer function. Nonetheless, RM can substantially alter the noise distribution. In this work, we utilize noise propagation models in order to accurately characterize the noise texture of reconstructed images in the presence of RM. Furthermore we consider the task of lesion or defect detection, which is highly determined by the noise distribution as quantified using the noise power spectrum. Ultimately, we use this framework to demonstrate why conventional trade-off analyses (e.g. contrast versus noise, using simplistic noise metrics) do not provide a complete picture of the impact of RM and that improved performance of RM according to such analyses does not necessarily translate to the superiority of RM in detection task performance.

Resolution modeling in PET imaging: theory, practice, benefits, and pitfalls

In this paper, the authors review the field of resolution modeling in positron emission tomography (PET) image reconstruction, also referred to as point-spread-function modeling. The review includes theoretical analysis of the resolution modeling framework as well as an overview of various approaches in the literature. It also discusses potential advantages gained via this approach, as discussed with reference to various metrics and tasks, including lesion detection observer studies. Furthermore, attention is paid to issues arising from this approach including the pervasive problem of edge artifacts, as well as explanation and potential remedies for this phenomenon. Furthermore, the authors emphasize limitations encountered in the context of quantitative PET imaging, wherein increased intervoxel correlations due to resolution modeling can lead to significant loss of precision (reproducibility) for small regions of interest, which can be a considerable pitfall depending on the task of interest.

Positron emission tomography imaging of fibrillar parenchymal and vascular amyloid-β in TgCRND8 mice

Few quantitative diagnostic and monitoring, tools are available to clinicians treating patients with Alzheimer's disease. Further, many of the promising quantitative imaging tools under development lack clear specificity toward different types of Amyloid-β (Aβ) pathology such as vascular or oligomeric species. Antibodies offer an opportunity to image specific types of Aβ pathology because of their excellent specificity. In this study, we developed a method to translate a panel of anti-Aβ antibodies, which show excellent histological performance, into live animal imaging contrast agents. In the TgCRND8 mouse model of Alzheimer's disease, we tested two antibodies, M64 and M116, that target parenchyma aggregated Aβ plaques and one antibody, M31, that targets vascular Aβ. All three antibodies were administered intravenously after labeling with both poly(ethylene glycol) to enhance circulation and (64)Cu to allow detection via positron emission tomography (PET) imaging. We were clearly able to differentiate TgCRND8 mice from wild type controls by PET imaging using either M116, the anti-Aβ antibody targeting parenchymal Aβ or M31, the antivascular Aβ antibody. To confirm the validity of the noninvasive imaging of specific Aβ pathology, brains were examined after imaging and showed clear evidence of binding to Aβ plaques.

Quantification of opioid receptor availability following spontaneous epileptic seizures: correction of [11C]diprenorphine PET data for the partial-volume effect

Previous positron emission tomography (PET) studies in refractory temporal lobe epilepsy (TLE) using the non-selective opioid receptor antagonist [(11)C]diprenorphine (DPN) did not detect any changes in mesial temporal structures, despite known involvement of the hippocampus in seizure generation. Normal binding in smaller hippocampi is suggestive of increased receptor concentration in the remaining grey matter. Correction for partial-volume effect (PVE) has not been used in previous DPN PET studies. Here, we present PVE-corrected DPN-PET data quantifying post-ictal and interictal opioid receptor availability in humans with mTLE. Eight paired datasets of post-ictal and interictal DPN PET scans and eleven test/retest control datasets were available from a previously published study on opioid receptor changes in TLE following seizures (Hammers et al., 2007a). Five of the eight participants with TLE had documented hippocampal sclerosis. Data were re-analyzed using regions of interest and a novel PVE correction method (structural functional synergistic-resolution recovery (SFS-RR); (Shidahara et al., 2012)). Data were denoised, followed by application of SFS-RR, with anatomical information derived via precise anatomical segmentation of the participants' MRI (MAPER; (Heckemann et al., 2010)). [(11)C]diprenorphine volume-of-distribution (VT) was quantified in six regions of interest. Post-ictal increases were observed in the ipsilateral fusiform gyri and lateral temporal pole. A novel finding was a post-ictal increase in [(11)C]DPN VT relative to the interictal state in the ipsilateral parahippocampal gyrus, not observed in uncorrected datasets. As for voxel-based (SPM) analyses, correction for global VT values was essential in order to demonstrate focal post-ictal increases in [(11)C]DPN VT. This study provides further direct human in vivo evidence for changes in opioid receptor availability in TLE following seizures, including changes that were not evident without PVE correction. Denoising, resolution recovery and precise anatomical segmentation can extract valuable information from PET studies that would be missed with conventional post-processing procedures.