Advancing Tau PET Quantification in Alzheimer Disease with Machine Learning: Introducing THETA, a Novel Tau Summary Measure.

Alzheimer disease (AD) exhibits spatially heterogeneous 3- or 4-repeat tau deposition across participants. Our overall goal was to develop an automated method to quantify the heterogeneous burden of tau deposition into a single number that would be clinically useful. Methods: We used tau PET scans from 3 independent cohorts: the Mayo Clinic Study of Aging and Alzheimer's Disease Research Center (Mayo, n = 1,290), the Alzheimer's Disease Neuroimaging Initiative (ADNI, n = 831), and the Open Access Series of Imaging Studies (OASIS-3, n = 430). A machine learning binary classification model was trained on Mayo data and validated on ADNI and OASIS-3 with the goal of predicting visual tau positivity (as determined by 3 raters following Food and Drug Administration criteria for 18F-flortaucipir). The machine learning model used region-specific SUV ratios scaled to cerebellar crus uptake. We estimated feature contributions based on an artificial intelligence-explainable method (Shapley additive explanations) and formulated a global tau summary measure, Tau Heterogeneity Evaluation in Alzheimer's Disease (THETA) score, using SUV ratios and Shapley additive explanations for each participant. We compared the performance of THETA with that of commonly used meta-regions of interest (ROIs) using the Mini-Mental State Examination, the Clinical Dementia Rating-Sum of Boxes, clinical diagnosis, and histopathologic staging. Results: The model achieved a balanced accuracy of 95% on the Mayo test set and at least 87% on the validation sets. It classified tau-positive and -negative participants with an AUC of 1.00, 0.96, and 0.94 on the Mayo, ADNI, and OASIS-3 cohorts, respectively. Across all cohorts, THETA showed a better correlation with the Mini-Mental State Examination and the Clinical Dementia Rating-Sum of Boxes (ρ ≥ 0.45, P < 0.05) than did meta-ROIs (ρ < 0.44, P < 0.05) and discriminated between participants who were cognitively unimpaired and those who had mild cognitive impairment with an effect size of 10.09, compared with an effect size of 3.08 for meta-ROIs. Conclusion: Our proposed approach identifies positive tau PET scans and provides a quantitative summary measure, THETA, that effectively captures heterogeneous tau deposition observed in AD. The application of THETA for quantifying tau PET in AD exhibits great potential.

Stress testing the Centiloid: Precision and variability of PET quantification of amyloid pathology.

INTRODUCTION: Assessing the potential sources of bias and variability of the Centiloid (CL) scale is fundamental for its appropriate clinical application. METHODS: We included 533 participants from AMYloid imaging to Prevent Alzheimer's Disease (AMYPAD DPMS) and Alzheimer's Disease Neuroimaging Initiative (ADNI) cohorts. Thirty-two CL pipelines were created using different combinations of reference region (RR), RR and target types, and quantification spaces. Generalized estimating equations stratified by amyloid positivity were used to assess the impact of the quantification pipeline, radiotracer, age, brain atrophy, and harmonization status on CL. RESULTS: RR selection and RR type impact CL the most, particularly in amyloid-negative individuals. The standard CL pipeline with the whole cerebellum as RR is robust against brain atrophy and differences in image resolution, with 95% confidence intervals below ± 3.95 CL for amyloid beta positivity cutoffs (CL < 24). DISCUSSION: The standard CL pipeline is recommended for most scenarios. Confidence intervals should be considered when operationalizing CL cutoffs in clinical and research settings. HIGHLIGHTS: We developed a framework for evaluating Centiloid (CL) variability to different factors. Reference region selection and delineation had the highest impact on CL values. Whole cerebellum (WCB) and whole cerebellum plus brainstem (WCB+BSTM) as reference regions yielded consistent results across tracers. The standard CL pipeline is robust against atrophy and image resolution variation. Estimated within- and between-pipeline variability (95% confidence interval) in absolute CL units.

Impact of simulated reduced injected dose on the assessment of amyloid PET scans.

PURPOSE: To investigate the impact of reduced injected doses on the quantitative and qualitative assessment of the amyloid PET tracers [18F]flutemetamol and [18F]florbetaben. METHODS: Cognitively impaired and unimpaired individuals (N = 250, 36% Aß-positive) were included and injected with [18F]flutemetamol (N = 175) or [18F]florbetaben (N = 75). PET scans were acquired in list-mode (90-110 min post-injection) and reduced-dose images were simulated to generate images of 75, 50, 25, 12.5 and 5% of the original injected dose. Images were reconstructed using vendor-provided reconstruction tools and visually assessed for Aß-pathology. SUVRs were calculated for a global cortical and three smaller regions using a cerebellar cortex reference tissue, and Centiloid was computed. Absolute and percentage differences in SUVR and CL were calculated between dose levels, and the ability to discriminate between Aß- and Aß + scans was evaluated using ROC analyses. Finally, intra-reader agreement between the reduced dose and 100% images was evaluated. RESULTS: At 5% injected dose, change in SUVR was 3.72% and 3.12%, with absolute change in Centiloid 3.35CL and 4.62CL, for [18F]flutemetamol and [18F]florbetaben, respectively. At 12.5% injected dose, percentage change in SUVR and absolute change in Centiloid were < 1.5%. AUCs for discriminating Aß- from Aß + scans were high (AUC ≥ 0.94) across dose levels, and visual assessment showed intra-reader agreement of > 80% for both tracers. CONCLUSION: This proof-of-concept study showed that for both [18F]flutemetamol and [18F]florbetaben, adequate quantitative and qualitative assessments can be obtained at 12.5% of the original injected dose. However, decisions to reduce the injected dose should be made considering the specific clinical or research circumstances.

Advancing Tau-PET quantification in Alzheimer's disease with machine learning: introducing THETA, a novel tau summary measure.

Alzheimer's disease (AD) exhibits spatially heterogeneous 3R/4R tau pathology distributions across participants, making it a challenge to quantify extent of tau deposition. Utilizing Tau-PET from three independent cohorts, we trained and validated a machine learning model to identify visually positive Tau-PET scans from regional SUVR values and developed a novel summary measure, THETA, that accounts for heterogeneity in tau deposition. The model for identification of tau positivity achieved a balanced test accuracy of 95% and accuracy of ≥87% on the validation datasets. THETA captured heterogeneity of tau deposition, had better association with clinical measures, and corresponded better with visual assessments in comparison with the temporal meta-region-of-interest Tau-PET quantification methods. Our novel approach aids in identification of positive Tau-PET scans and provides a quantitative summary measure, THETA, that effectively captures the heterogeneous tau deposition seen in AD. The application of THETA for quantifying Tau-PET in AD exhibits great potential.

Genetically identical twin-pair difference models support the amyloid cascade hypothesis.

The amyloid cascade hypothesis has strongly impacted the Alzheimer's disease research agenda and clinical trial designs over the past decades, but precisely how amyloid-ß pathology initiates the aggregation of neocortical tau remains unclear. We cannot exclude the possibility of a shared upstream process driving both amyloid-ß and tau in an independent manner instead of there being a causal relationship between amyloid-ß and tau. Here, we tested the premise that if a causal relationship exists, then exposure should be associated with outcome both at the individual level as well as within identical twin-pairs, who are strongly matched on genetic, demographic and shared environmental background. Specifically, we tested associations between longitudinal amyloid-ß PET and cross-sectional tau PET, neurodegeneration and cognitive decline using genetically identical twin-pair difference models, which provide the unique opportunity of ruling out genetic and shared environmental effects as potential confounders in an association. We included 78 cognitively unimpaired identical twins with [18F]flutemetamol (amyloid-ß)-PET, [18F]flortaucipir (tau)-PET, MRI (hippocampal volume) and cognitive data (composite memory). Associations between each modality were tested at the individual level using generalized estimating equation models, and within identical twin-pairs using within-pair difference models. Mediation analyses were performed to test for directionality in the associations as suggested by the amyloid cascade hypothesis. At the individual level, we observed moderate-to-strong associations between amyloid-ß, tau, neurodegeneration and cognition. The within-pair difference models replicated results observed at the individual level with comparably strong effect sizes. Within-pair differences in amyloid-ß were strongly associated with within-pair differences in tau (ß = 0.68, P < 0.001), and moderately associated with within-pair differences in hippocampal volume (ß = -0.37, P = 0.03) and memory functioning (ß = -0.57, P < 0.001). Within-pair differences in tau were moderately associated with within-pair differences in hippocampal volume (ß = -0.53, P < 0.001) and strongly associated with within-pair differences in memory functioning (ß = -0.68, P < 0.001). Mediation analyses showed that of the total twin-difference effect of amyloid-ß on memory functioning, the proportion mediated through pathways including tau and hippocampal volume was 69.9%, which was largely attributable to the pathway leading from amyloid-ß to tau to memory functioning (proportion mediated, 51.6%). Our results indicate that associations between amyloid-ß, tau, neurodegeneration and cognition are unbiased by (genetic) confounding. Furthermore, effects of amyloid-ß on neurodegeneration and cognitive decline were fully mediated by tau. These novel findings in this unique sample of identical twins are compatible with the amyloid cascade hypothesis and thereby provide important new knowledge for clinical trial designs.

NiftyPAD - Novel Python Package for Quantitative Analysis of Dynamic PET Data.

Current PET datasets are becoming larger, thereby increasing the demand for fast and reproducible processing pipelines. This paper presents a freely available, open source, Python-based software package called NiftyPAD, for versatile analyses of static, full or dual-time window dynamic brain PET data. The key novelties of NiftyPAD are the analyses of dual-time window scans with reference input processing, pharmacokinetic modelling with shortened PET acquisitions through the incorporation of arterial spin labelling (ASL)-derived relative perfusion measures, as well as optional PET data-based motion correction. Results obtained with NiftyPAD were compared with the well-established software packages PPET and QModeling for a range of kinetic models. Clinical data from eight subjects scanned with four different amyloid tracers were used to validate the computational performance. NiftyPAD achieved [Formula: see text] correlation with PPET, with absolute difference [Formula: see text] for linearised Logan and MRTM2 methods, and [Formula: see text] correlation with QModeling, with absolute difference [Formula: see text] for basis function based SRTM and SRTM2 models. For the recently published SRTM ASL method, which is unavailable in existing software packages, high correlations with negligible bias were observed with the full scan SRTM in terms of non-displaceable binding potential ([Formula: see text]), indicating reliable model implementation in NiftyPAD. Together, these findings illustrate that NiftyPAD is versatile, flexible, and produces comparable results with established software packages for quantification of dynamic PET data. It is freely available ( https://github.com/AMYPAD/NiftyPAD ), and allows for multi-platform usage. The modular setup makes adding new functionalities easy, and the package is lightweight with minimal dependencies, making it easy to use and integrate into existing processing pipelines.

Precision estimates of relative and absolute cerebral blood flow in Alzheimer's disease and cognitively normal individuals.

Alzheimer's disease is characterized by regional reductions in cerebral blood flow (CBF). Although the gold standard for measuring CBF is [15O]H2O PET, proxies of relative CBF, derived from the early distribution phase of amyloid and tau tracers, have gained attention. The present study assessed precision of [15O]H2O derived relative and absolute CBF, and compared precision of these measures with that of (relative) CBF proxies. Dynamic [15O]H2O, [18F]florbetapir and [18F]flortaucipir PET test-retest (TrT) datasets with eleven, nine and fourteen subjects, respectively, were included. Analyses were performed using an arterial input model and/or a simplified reference tissue model, depending on the data available. Relative CBF values (i.e. K1/K1' and/or R1) were obtained using cerebellar cortex as reference tissue and TrT repeatability (i.e. precision) was calculated and compared between tracers, parameters and clinical groups. Relative CBF had significantly better TrT repeatability than absolute CBF derived from [15O]H2O (r = -0.53), while best TrT repeatability was observed for [18F]florbetapir and [18F]flortaucipir R1 (r = -0.23, r = -0.33). Furthermore, only R1 showed, better TrT repeatability for cognitively normal individuals. High precision of CBF proxies could be due to a compensatory effect of the extraction fraction, although changes in extraction fraction could also bias these proxies, but not the gold standard.

Impact of cerebral blood flow and amyloid load on SUVR bias.

BACKGROUND: Despite its widespread use, the semi-quantitative standardized uptake value ratio (SUVR) may be biased compared with the distribution volume ratio (DVR). This bias may be partially explained by changes in cerebral blood flow (CBF) and is likely to be also dependent on the extent of the underlying amyloid-ß (Aß) burden. This study aimed to compare SUVR with DVR and to evaluate the effects of underlying Aß burden and CBF on bias in SUVR in mainly cognitively unimpaired participants. Participants were scanned according to a dual-time window protocol, with either [18F]flutemetamol (N = 90) or [18F]florbetaben (N = 31). The validated basisfunction-based implementation of the two-step simplified reference tissue model was used to derive DVR and R1 parametric images, and SUVR was calculated from 90 to 110 min post-injection, all with the cerebellar grey matter as reference tissue. First, linear regression and Bland-Altman analyses were used to compare (regional) SUVR with DVR. Then, generalized linear models were applied to evaluate whether (bias in) SUVR relative to DVR could be explained by R1 for the global cortical average (GCA), precuneus, posterior cingulate, and orbitofrontal region. RESULTS: Despite high correlations (GCA: R2 ≥ 0.85), large overestimation and proportional bias of SUVR relative to DVR was observed. Negative associations were observed between both SUVR or SUVRbias and R1, albeit non-significant. CONCLUSION: The present findings demonstrate that bias in SUVR relative to DVR is strongly related to underlying Aß burden. Furthermore, in a cohort consisting mainly of cognitively unimpaired individuals, the effect of relative CBF on bias in SUVR appears limited. EudraCT Number: 2018-002277-22, registered on: 25-06-2018.

Quantification of amyloid PET for future clinical use: a state-of-the-art review.

Amyloid-ß (Aß) pathology is one of the earliest detectable brain changes in Alzheimer's disease (AD) pathogenesis. The overall load and spatial distribution of brain Aß can be determined in vivo using positron emission tomography (PET), for which three fluorine-18 labelled radiotracers have been approved for clinical use. In clinical practice, trained readers will categorise scans as either Aß positive or negative, based on visual inspection. Diagnostic decisions are often based on these reads and patient selection for clinical trials is increasingly guided by amyloid status. However, tracer deposition in the grey matter as a function of amyloid load is an inherently continuous process, which is not sufficiently appreciated through binary cut-offs alone. State-of-the-art methods for amyloid PET quantification can generate tracer-independent measures of Aß burden. Recent research has shown the ability of these quantitative measures to highlight pathological changes at the earliest stages of the AD continuum and generate more sensitive thresholds, as well as improving diagnostic confidence around established binary cut-offs. With the recent FDA approval of aducanumab and more candidate drugs on the horizon, early identification of amyloid burden using quantitative measures is critical for enrolling appropriate subjects to help establish the optimal window for therapeutic intervention and secondary prevention. In addition, quantitative amyloid measurements are used for treatment response monitoring in clinical trials. In clinical settings, large multi-centre studies have shown that amyloid PET results change both diagnosis and patient management and that quantification can accurately predict rates of cognitive decline. Whether these changes in management reflect an improvement in clinical outcomes is yet to be determined and further validation work is required to establish the utility of quantification for supporting treatment endpoint decisions. In this state-of-the-art review, several tools and measures available for amyloid PET quantification are summarised and discussed. Use of these methods is growing both clinically and in the research domain. Concurrently, there is a duty of care to the wider dementia community to increase visibility and understanding of these methods.

Spatial-Temporal Patterns of ß-Amyloid Accumulation: A Subtype and Stage Inference Model Analysis.

BACKGROUND AND OBJECTIVES: ß-amyloid (Aß) staging models assume a single spatial-temporal progression of amyloid accumulation. We assessed evidence for Aß accumulation subtypes by applying the data-driven Subtype and Stage Inference (SuStaIn) model to amyloid-PET data. METHODS: Amyloid-PET data of 3,010 participants were pooled from 6 cohorts (ALFA+, EMIF-AD, ABIDE, OASIS, and ADNI). Standardized uptake value ratios were calculated for 17 regions. We applied the SuStaIn algorithm to identify consistent subtypes in the pooled dataset based on the cross-validation information criterion and the most probable subtype/stage classification per scan. The effects of demographics and risk factors on subtype assignment were assessed using multinomial logistic regression. RESULTS: Participants were mostly cognitively unimpaired (n = 1890 [62.8%]), had a mean age of 68.72 (SD 9.1) years, 42.1% were APOE ε4 carriers, and 51.8% were female. A 1-subtype model recovered the traditional amyloid accumulation trajectory, but SuStaIn identified 3 optimal subtypes, referred to as frontal, parietal, and occipital based on the first regions to show abnormality. Of the 788 (26.2%) with strong subtype assignment (>50% probability), the majority was assigned to frontal (n = 415 [52.5%]), followed by parietal (n = 199 [25.3%]) and occipital subtypes (n = 175 [22.2%]). Significant differences across subtypes included distinct proportions of APOE ε4 carriers (frontal 61.8%, parietal 57.1%, occipital 49.4%), participants with dementia (frontal 19.7%, parietal 19.1%, occipital 31.0%), and lower age for the parietal subtype (frontal/occipital 72.1 years, parietal 69.3 years). Higher amyloid (Centiloid) and CSF p-tau burden was observed for the frontal subtype; parietal and occipital subtypes did not differ. At follow-up, most participants (81.1%) maintained baseline subtype assignment and 25.6% progressed to a later stage. DISCUSSION: Whereas a 1-trajectory model recovers the established pattern of amyloid accumulation, SuStaIn determined that 3 subtypes were optimal, showing distinct associations with Alzheimer disease risk factors. Further analyses to determine clinical utility are warranted.

Test-Retest Variability of Relative Tracer Delivery Rate as Measured by [11C]PiB.

PURPOSE: Moderate-to-high correlations have been reported between the [11C]PiB PET-derived relative tracer delivery rate R1 and relative CBF as measured using [15O]H2O PET, supporting its use as a proxy of relative CBF. As longitudinal PET studies become more common for measuring treatment efficacy or disease progression, it is important to know the intrinsic variability of R1. The purpose of the present study was to determine this through a retrospective data analysis. PROCEDURES: Test-retest data belonging to twelve participants, who underwent two 90 min [11C]PiB PET scans, were retrospectively included. The voxel-based implementation of the two-step simplified reference tissue model with cerebellar grey matter as reference tissue was used to compute R1 images. Next, test-retest variability was calculated, and test and retest R1 measures were compared using linear mixed effect models and a Bland-Altman analysis. RESULTS: Test-retest variability was low across regions (max. 5.8 %), and test and retest measures showed high, significant correlations (R2=0.92, slope=0.98) and a negligible bias (0.69±3.07 %). CONCLUSIONS: In conclusion, the high precision of [11C]PiB R1 suggests suitable applicability for cross-sectional and longitudinal studies.

Strategies to reduce sample sizes in Alzheimer's disease primary and secondary prevention trials using longitudinal amyloid PET imaging.

BACKGROUND: Detecting subtle-to-moderate biomarker changes such as those in amyloid PET imaging becomes increasingly relevant in the context of primary and secondary prevention of Alzheimer's disease (AD). This work aimed to determine if and when distribution volume ratio (DVR; derived from dynamic imaging) and regional quantitative values could improve statistical power in AD prevention trials. METHODS: Baseline and annualized % change in [11C]PIB SUVR and DVR were computed for a global (cortical) and regional (early) composite from scans of 237 cognitively unimpaired subjects from the OASIS-3 database ( www.oasis-brains.org ). Bland-Altman and correlation analyses were used to assess the relationship between SUVR and DVR. General linear models and linear mixed effects models were used to determine effects of age, sex, and APOE-ε4 carriership on baseline and longitudinal amyloid burden. Finally, differences in statistical power of SUVR and DVR (cortical or early composite) were assessed considering three anti-amyloid trial scenarios: secondary prevention trials including subjects with (1) intermediate-to-high (Centiloid > 20.1), or (2) intermediate (20.1 < Centiloid ≤ 49.4) amyloid burden, and (3) a primary prevention trial focusing on subjects with low amyloid burden (Centiloid ≤ 20.1). Trial scenarios were set to detect 20% reduction in accumulation rates across the whole population and in APOE-ε4 carriers only. RESULTS: Although highly correlated to DVR (ρ = .96), cortical SUVR overestimated DVR cross-sectionally and in annual % change. In secondary prevention trials, DVR required 143 subjects per arm, compared with 176 for SUVR. Both restricting inclusion to individuals with intermediate amyloid burden levels or to APOE-ε4 carriers alone further reduced sample sizes. For primary prevention, SUVR required less subjects per arm (n = 855) compared with DVR (n = 1508) and the early composite also provided considerable sample size reductions (n = 855 to n = 509 for SUVR, n = 1508 to n = 734 for DVR). CONCLUSION: Sample sizes in AD secondary prevention trials can be reduced by the acquisition of dynamic PET scans and/or by restricting inclusion to subjects with intermediate amyloid burden or to APOE-ε4 carriers only. Using a targeted early composite only leads to reductions of sample size requirements in primary prevention trials. These findings support strategies to enable smaller Proof-of-Concept Phase II clinical trials to better streamline drug development.

Parametric imaging of dual-time window [18F]flutemetamol and [18F]florbetaben studies.

Optimal pharmacokinetic models for quantifying amyloid beta (Aß) burden using both [18F]flutemetamol and [18F]florbetaben scans have previously been identified at a region of interest (ROI) level. The purpose of this study was to determine optimal quantitative methods for parametric analyses of [18F]flutemetamol and [18F]florbetaben scans. Forty-six participants were scanned on a PET/MR scanner using a dual-time window protocol and either [18F]flutemetamol (N=24) or [18F]florbetaben (N=22). The following parametric approaches were used to derive DVR estimates: reference Logan (RLogan), receptor parametric mapping (RPM), two-step simplified reference tissue model (SRTM2) and multilinear reference tissue models (MRTM0, MRTM1, MRTM2), all with cerebellar grey matter as reference tissue. In addition, a standardized uptake value ratio (SUVR) was calculated for the 90-110 min post injection interval. All parametric images were assessed visually. Regional outcome measures were compared with those from a validated ROI method, i.e. DVR derived using RLogan. Visually, RPM, and SRTM2 performed best across tracers and, in addition to SUVR, provided highest AUC values for differentiating between Aß-positive vs Aß-negative scans ([18F]flutemetamol: range AUC=0.96-0.97 [18F]florbetaben: range AUC=0.83-0.85). Outcome parameters of most methods were highly correlated with the reference method (R2≥0.87), while lowest correlation were observed for MRTM2 (R2=0.71-0.80). Furthermore, bias was low (≤5%) and independent of underlying amyloid burden for MRTM0 and MRTM1. The optimal parametric method differed per evaluated aspect; however, the best compromise across aspects was found for MRTM0 followed by SRTM2, for both tracers. SRTM2 is the preferred method for parametric imaging because, in addition to its good performance, it has the advantage of providing a measure of relative perfusion (R1), which is useful for measuring disease progression.

Simulating the effect of cerebral blood flow changes on regional quantification of [18F]flutemetamol and [18F]florbetaben studies.

Global and regional changes in cerebral blood flow (CBF) can result in biased quantitative estimates of amyloid load by PET imaging. Therefore, the current simulation study assessed effects of these changes on amyloid quantification using a reference tissue approach for [18F]flutemetamol and [18F]florbetaben. Previously validated pharmacokinetic rate constants were used to simulate time-activity curves (TACs) corresponding to full dynamic and dual-time-window acquisition protocols. CBF changes were simulated by varying the tracer delivery (K1) from +25 to -25%. The standardized uptake value ratio (SUVr) was computed and TACs were fitted using reference Logan (RLogan) and the simplified reference tissue model (SRTM) to obtain the relative delivery rate (R1) and volume of distribution ratio (DVR). RLogan was least affected by CBF changes (χ2 = 583 p < 0.001, χ2 = 81 p < 0.001, for [18F]flutemetamol and [18F]florbetaben, respectively) and the extent of CBF sensitivity generally increased for higher levels of amyloid. Further, SRTM-derived R1 changes correlated well with simulated CBF changes (R2 > 0.95) and SUVr's sensitivity to CBF changes improved for later uptake-times, with the exception of [18F]flutemetamol cortical changes. In conclusion, RLogan is the preferred method for amyloid quantification of [18F]flutemetamol and [18F]florbetaben studies and SRTM could be additionally used for obtaining a CBF proxy.

Classification of negative and positive 18F-florbetapir brain PET studies in subjective cognitive decline patients using a convolutional neural network.

PURPOSE: Visual reading of 18F-florbetapir positron emission tomography (PET) scans is used in the diagnostic process of patients with cognitive disorders for assessment of amyloid-ß (Aß) depositions. However, this can be time-consuming, and difficult in case of borderline amyloid pathology. Computer-aided pattern recognition can be helpful in this process but needs to be validated. The aim of this work was to develop, train, validate and test a convolutional neural network (CNN) for discriminating between Aß negative and positive 18F-florbetapir PET scans in patients with subjective cognitive decline (SCD). METHODS: 18F-florbetapir PET images were acquired and visually assessed. The SCD cohort consisted of 133 patients from the SCIENCe cohort and 22 patients from the ADNI database. From the SCIENCe cohort, standardized uptake value ratio (SUVR) images were computed. From the ADNI database, SUVR images were extracted. 2D CNNs (axial, coronal and sagittal) were built to capture features of the scans. The SCIENCe scans were randomly divided into training and validation set (5-fold cross-validation), and the ADNI scans were used as test set. Performance was evaluated based on average accuracy, sensitivity and specificity from the cross-validation. Next, the best performing CNN was evaluated on the test set. RESULTS: The sagittal 2D-CNN classified the SCIENCe scans with the highest average accuracy of 99% ± 2 (SD), sensitivity of 97% ± 7 and specificity of 100%. The ADNI scans were classified with a 95% accuracy, 100% sensitivity and 92.3% specificity. CONCLUSION: The 2D-CNN algorithm can classify Aß negative and positive 18F-florbetapir PET scans with high performance in SCD patients.

[11C]PIB amyloid quantification: effect of reference region selection.

BACKGROUND: The standard reference region (RR) for amyloid-beta (Aß) PET studies is the cerebellar grey matter (GMCB), while alternative RRs have mostly been utilized without prior validation against the gold standard. This study compared five commonly used RRs to gold standard plasma input-based quantification using the GMCB. METHODS: Thirteen subjects from a test-retest (TRT) study and 30 from a longitudinal study were retrospectively included (total: 17 Alzheimer's disease, 13 mild cognitive impairment, 13 controls). Dynamic [11C]PiB PET (90 min) and T1-weighted MR scans were co-registered and time-activity curves were extracted for cortical target regions and the following RRs: GMCB, whole cerebellum (WCB), white matter brainstem/pons (WMBS), whole brainstem (WBS) and eroded subcortical white matter (WMES). A two-tissue reversible plasma input model (2T4k_Vb) with GMCB as RR, reference Logan and the simplified reference tissue model were used to derive distribution volume ratios (DVRs), and standardized uptake value (SUV) ratios were calculated for 40-60 min and 60-90 min intervals. Parameter variability was evaluated using TRT scans, and correlations and agreements with the gold standard (DVR from 2T4k_Vb with GMCB RR) were also assessed. Next, longitudinal changes in SUVs (both intervals) were assessed for each RR. Finally, the ability to discriminate between visually Aß positive and Aß negative scans was assessed. RESULTS: All RRs yielded stable TRT performance (max 5.1% variability), with WCB consistently showing lower variability. All approaches were able to discriminate between Aß positive and Aß negative scans, with highest effect sizes obtained for GMCB (range - 0.9 to - 0.7), followed by WCB (range - 0.8 to - 0.6). Furthermore, all approaches provided good correlations with the gold standard (r ≥ 0.78), while the highest bias (as assessed by the regression slope) was observed using WMES (range slope 0.52-0.67), followed by WBS (range slope 0.58-0.92) and WMBS (range slope 0.62-0.91). Finally, RR SUVs were stable across a period of 2.6 years for all except WBS and WMBS RRs (60-90 min interval). CONCLUSIONS: GMCB and WCB are considered the best RRs for quantifying amyloid burden using [11C]PiB PET.

Multitracer model for staging cortical amyloid deposition using PET imaging.

OBJECTIVE: To develop and evaluate a model for staging cortical amyloid deposition using PET with high generalizability. METHODS: Three thousand twenty-seven individuals (1,763 cognitively unimpaired [CU], 658 impaired, 467 with Alzheimer disease [AD] dementia, 111 with non-AD dementia, and 28 with missing diagnosis) from 6 cohorts (European Medical Information Framework for AD, Alzheimer's and Family, Alzheimer's Biomarkers in Daily Practice, Amsterdam Dementia Cohort, Open Access Series of Imaging Studies [OASIS]-3, Alzheimer's Disease Neuroimaging Initiative [ADNI]) who underwent amyloid PET were retrospectively included; 1,049 individuals had follow-up scans. With application of dataset-specific cutoffs to global standard uptake value ratio (SUVr) values from 27 regions, single-tracer and pooled multitracer regional rankings were constructed from the frequency of abnormality across 400 CU individuals (100 per tracer). The pooled multitracer ranking was used to create a staging model consisting of 4 clusters of regions because it displayed a high and consistent correlation with each single-tracer ranking. Relationships between amyloid stage, clinical variables, and longitudinal cognitive decline were investigated. RESULTS: SUVr abnormality was most frequently observed in cingulate, followed by orbitofrontal, precuneal, and insular cortices and then the associative, temporal, and occipital regions. Abnormal amyloid levels based on binary global SUVr classification were observed in 1.0%, 5.5%, 17.9%, 90.0%, and 100.0% of individuals in stage 0 to 4, respectively. Baseline stage predicted decline in Mini-Mental State Examination (MMSE) score (ADNI: n = 867, F = 67.37, p < 0.001; OASIS: n = 475, F = 9.12, p < 0.001) and faster progression toward an MMSE score ≤25 (ADNI: n = 787, hazard ratio [HR]stage1 2.00, HRstage2 3.53, HRstage3 4.55, HRstage4 9.91, p < 0.001; OASIS: n = 469, HRstage4 4.80, p < 0.001). CONCLUSION: The pooled multitracer staging model successfully classified the level of amyloid burden in >3,000 individuals across cohorts and radiotracers and detects preglobal amyloid burden and distinct risk profiles of cognitive decline within globally amyloid-positive individuals.

Optimized dual-time-window protocols for quantitative [18F]flutemetamol and [18F]florbetaben PET studies.

BACKGROUND: A long dynamic scanning protocol may be required to accurately measure longitudinal changes in amyloid load. However, such a protocol results in a lower patient comfort and scanning efficiency compared to static scans. A compromise can be achieved by implementing dual-time-window protocols. This study aimed to optimize these protocols for quantitative [18F]flutemetamol and [18F]florbetaben studies. METHODS: Rate constants for subjects across the Alzheimer's disease spectrum (i.e., non-displaceable binding potential (BPND) in the range 0.02-0.77 and 0.02-1.04 for [18F]flutemetamol and [18F]florbetaben, respectively) were established based on clinical [18F]flutemetamol (N = 6) and [18F]florbetaben (N = 20) data, and used to simulate tissue time-activity curves (TACs) of 110 min using a reference tissue and plasma input model. Next, noise was added (N = 50) and data points corresponding to different intervals were removed from the TACs, ranging from 0 (i.e., 90-90 = full-kinetic curve) to 80 (i.e., 10-90) minutes, creating a dual-time-window. Resulting TACs were fitted using the simplified reference tissue method (SRTM) to estimate the BPND, outliers (≥ 1.5 × BPND max) were removed and the bias was assessed using the distribution volume ratio (DVR = BPND + 1). To this end, acceptability curves, which display the fraction of data below a certain bias threshold, were generated and the area under those curves were calculated. RESULTS: [18F]Flutemetamol and [18F]florbetaben data demonstrated an increased bias in amyloid estimate for larger intervals and higher noise levels. An acceptable bias (≤ 3.1%) in DVR could be obtained with all except the 10-90 and 20-90-min intervals. Furthermore, a reduced fraction of acceptable data and most outliers were present for these two largest intervals (maximum percentage outliers 48 and 32 for [18F]flutemetamol and [18F]florbetaben, respectively). CONCLUSIONS: The length of the interval inversely correlates with the accuracy of the BPND estimates. Consequently, a dual-time-window protocol of 0-30 and 90-110 min (=maximum of 60 min interval) allows for accurate estimation of BPND values for both tracers. [18F]flutemetamol: EudraCT 2007-000784-19, registered 8 February 2007, [18F]florbetaben: EudraCT 2006-003882-15, registered 2006.

Application of Machine Learning to Arterial Spin Labeling in Mild Cognitive Impairment and Alzheimer Disease.

Purpose To investigate whether multivariate pattern recognition analysis of arterial spin labeling (ASL) perfusion maps can be used for classification and single-subject prediction of patients with Alzheimer disease (AD) and mild cognitive impairment (MCI) and subjects with subjective cognitive decline (SCD) after using the W score method to remove confounding effects of sex and age. Materials and Methods Pseudocontinuous 3.0-T ASL images were acquired in 100 patients with probable AD; 60 patients with MCI, of whom 12 remained stable, 12 were converted to a diagnosis of AD, and 36 had no follow-up; 100 subjects with SCD; and 26 healthy control subjects. The AD, MCI, and SCD groups were divided into a sex- and age-matched training set (n = 130) and an independent prediction set (n = 130). Standardized perfusion scores adjusted for age and sex (W scores) were computed per voxel for each participant. Training of a support vector machine classifier was performed with diagnostic status and perfusion maps. Discrimination maps were extracted and used for single-subject classification in the prediction set. Prediction performance was assessed with receiver operating characteristic (ROC) analysis to generate an area under the ROC curve (AUC) and sensitivity and specificity distribution. Results Single-subject diagnosis in the prediction set by using the discrimination maps yielded excellent performance for AD versus SCD (AUC, 0.96; P < .01), good performance for AD versus MCI (AUC, 0.89; P < .01), and poor performance for MCI versus SCD (AUC, 0.63; P = .06). Application of the AD versus SCD discrimination map for prediction of MCI subgroups resulted in good performance for patients with MCI diagnosis converted to AD versus subjects with SCD (AUC, 0.84; P < .01) and fair performance for patients with MCI diagnosis converted to AD versus those with stable MCI (AUC, 0.71; P > .05). Conclusion With automated methods, age- and sex-adjusted ASL perfusion maps can be used to classify and predict diagnosis of AD, conversion of MCI to AD, stable MCI, and SCD with good to excellent accuracy and AUC values. © RSNA, 2016.