The prognostic value of amyloid imaging

Ethical issues when using digital biomarkers and artificial intelligence for the early detection of dementia

Dementia poses a growing challenge for health services but remains stigmatized and under-recognized. Digital technologies to aid the earlier detection of dementia are approaching market. These include traditional cognitive screening tools presented on mobile devices, smartphone native applications, passive data collection from wearable, in-home and in-car sensors, as well as machine learning techniques applied to clinic and imaging data. It has been suggested that earlier detection and diagnosis may help patients plan for their future, achieve a better quality of life, and access clinical trials and possible future disease modifying treatments. In this review, we explore whether digital tools for the early detection of dementia can or should be deployed, by assessing them against the principles of ethical screening programs. We conclude that while the importance of dementia as a health problem is unquestionable, significant challenges remain. There is no available treatment which improves the prognosis of diagnosed disease. Progression from early-stage disease to dementia is neither given nor currently predictable. Available technologies are generally not both minimally invasive and highly accurate. Digital deployment risks exacerbating health inequalities due to biased training data and inequity in digital access. Finally, the acceptability of early dementia detection is not established, and resources would be needed to ensure follow-up and support for those flagged by any new system. We conclude that early dementia detection deployed at scale via digital technologies does not meet standards for a screening program and we offer recommendations for moving toward an ethical mode of implementation. This article is categorized under:Application Areas > Health CareCommercial, Legal, and Ethical Issues > Ethical ConsiderationsTechnologies > Artificial Intelligence.

Predicting Brain Amyloid Positivity from T1 weighted brain MRI and MRI-derived Gray Matter, White Matter and CSF maps using Transfer Learning on 3D CNNs

Abnormal β-amyloid (Aβ) accumulation in the brain is an early indicator of Alzheimer's disease and practical tests could help identify patients who could respond to treatment, now that promising anti-amyloid drugs are available. Even so, Aβ positivity (Aβ+) is assessed using PET or CSF assays, both highly invasive procedures. Here, we investigate how well Aβ+ can be predicted from T1 weighted brain MRI and gray matter, white matter and cerebrospinal fluid segmentations from T1-weighted brain MRI (T1w), a less invasive alternative. We used 3D convolutional neural networks to predict Aβ+ based on 3D brain MRI data, from 762 elderly subjects (mean age: 75.1 yrs. ± 7.6SD; 394F/368M; 459 healthy controls, 67 with MCI and 236 with dementia) scanned as part of the Alzheimer's Disease Neuroimaging Initiative. We also tested whether the accuracy increases when using transfer learning from the larger UK Biobank dataset. Overall, the 3D CNN predicted Aβ+ with 76% balanced accuracy from T1w scans. The closest performance to this was using white matter maps alone when the model was pre-trained on an age prediction in the UK Biobank. The performance of individual tissue maps was less than the T1w, but transfer learning helped increase the accuracy. Although tests on more diverse data are warranted, deep learned models from standard MRI show initial promise for Aβ+ estimation, before considering more invasive procedures.

Machine Learning for the Prediction of Amyloid Positivity in Amnestic Mild Cognitive Impairment

BACKGROUND

Amyloid-β (Aβ) evaluation in amnestic mild cognitive impairment (aMCI) patients is important for predicting conversion to Alzheimer's disease. However, Aβ evaluation through Aβ positron emission tomography (PET) is limited due to high cost and safety issues.

OBJECTIVE

We therefore aimed to develop and validate prediction models of Aβ positivity for aMCI using optimal interpretable machine learning (ML) approaches utilizing multimodal markers.

METHODS

We recruited 529 aMCI patients from multiple centers who underwent Aβ PET. We trained ML algorithms using a training cohort (324 aMCI from Samsung medical center) with two-phase modelling: model 1 included age, gender, education, diabetes, hypertension, apolipoprotein E genotype, and neuropsychological test scores; model 2 included the same variables as model 1 with additional MRI features. We used four-fold cross-validation during the modelling and evaluated the models on an external validation cohort (187 aMCI from the other centers).

RESULTS

Model 1 showed good accuracy (area under the receiver operating characteristic curve [AUROC] 0.837) in cross-validation, and fair accuracy (AUROC 0.765) in external validation. Model 2 led to improvement in the prediction performance with good accuracy (AUROC 0.892) in cross validation compared to model 1. Apolipoprotein E genotype, delayed recall task scores, and interaction between cortical thickness in the temporal region and hippocampal volume were the most important predictors of Aβ positivity.

CONCLUSION

Our results suggest that ML models are effective in predicting Aβ positivity at the individual level and could help the biomarker-guided diagnosis of prodromal AD.

A Nomogram for Predicting Amyloid PET Positivity in Amnestic Mild Cognitive Impairment

BACKGROUND

Most clinical trials focus on amyloid-β positive (Aβ+) amnestic mild cognitive impairment (aMCI), but screening failures are high because only a half of patients with aMCI are positive on Aβ PET. Therefore, it becomes necessary for clinicians to predict which patients will have Aβ biomarker.

OBJECTIVE

We aimed to compare clinical factors, neuropsychological (NP) profiles, and apolipoprotein E (APOE) genotype between Aβ+ aMCI and Aβ-aMCI and to develop a clinically useful prediction model of Aβ positivity on PET (PET-Aβ+) in aMCI using a nomogram.

METHODS

We recruited 523 aMCI patients who underwent Aβ PET imaging in a nation-wide multicenter cohort. The results of NP measures were divided into following subgroups: 1) Stage (Early and Late-stage), 2) Modality (Visual, Verbal, and Both), 3) Recognition failure, and 4) Multiplicity (Single and Multiple). A nomogram for PET-Aβ+ in aMCI patients was constructed using a logistic regression model.

RESULTS

PET-Aβ+ had significant associations with NP profiles for several items, including high Clinical Dementia Rating Scale Sum of Boxes score (OR 1.47, p = 0.013) and impaired memory modality (impaired both visual and verbal memories compared with visual only, OR 3.25, p = 0.001). Also, presence of APOEɛ4 (OR 4.14, p < 0.001) was associated with PET-Aβ+. These predictors were applied to develop the nomogram, which showed good prediction performance (C-statistics = 0.79). Its prediction performances were 0.77/0.74 in internal/external validation.

CONCLUSIONS

The nomogram consisting of NP profiles, especially memory domain, and APOEɛ4 genotype may provide a useful predictive model of PET-Aβ+ in patients with aMCI.

Negative 11C-PIB PET Predicts Lack of Alzheimer's Disease Pathology in Postmortem Examination

Our aim was to assess whether in vivo11C-PIB negative memory-impaired subjects may nonetheless exhibit brain Alzheimer's disease (AD) pathology. We re-evaluated the PET images and systematically characterized the postmortem neuropathology of six individuals who had undergone clinically indicated amyloid PET. The single case with negligible amyloid-β (Aβ) pathology had Lewy body disease, where concomitant AD changes are often seen. Further, the subject's plaques were predominantly diffuse. The predictive value of a negative 11C-PIB scan appears to be good, even in memory-impaired populations. Our results suggest that considerable neuritic Aβ plaque pathology in the absence of specific/cortical 11C-PIB binding upon PET is unlikely.

Search for Alzheimer's disease biomarkers in blood cells: hypotheses-driven approach

Current Alzheimer's disease (AD) diagnostics is based on cognitive testing, and detecting amyloid Aβ and τ pathology by brain imaging and assays of cerebrospinal fluid. However, biomarkers identifying complex pathways contributing to pathology are lacking, especially for early AD. Preferably, such biomarkers should be more cost-effective and present in easily available diagnostic tissues, such as blood. Here, we summarize the recent findings of potential early AD molecular diagnostic biomarkers in blood platelets, lymphocytes and erythrocytes. We review molecular alterations which refer to such main hypotheses of AD pathogenesis as amyloid cascade, oxidative and mitochondrial stress, inflammation and alterations in cell cycle regulatory molecules. The major advantage of such biomarkers is the potential ability to indicate individualized therapies in AD patients.

Profile of 6 microRNA in blood plasma distinguish early stage Alzheimer's disease patients from non-demented subjects

Alzheimer's disease (AD) is the most common age-related dementia. Among its major challenges is identifying molecular signatures characteristic for the early AD stage in patients with Mild Cognitive Impairment (MCI-AD), which could serve for deciphering the AD pathomechanism and also as non-invasive, easy-to-access biomarkers. Using qRT-PCR we compared the microRNA (miRNA) profiles in blood plasma of 15 MCI-AD patients, whose diagnoses were confirmed by cerebrospinal fluid (CSF) biomarkers, with 20 AD patients and 15 non-demented, age-matched individuals (CTR).To minimize methodological variability, we adhered to standardization of blood and CSF assays recommended by the international Joint Programming for Neurodegenerative Diseases (JPND) BIOMARKAPD consortium, and we employed commercially available Exiqon qRT-PCR-assays. In the first screening, we assessed 179 miRNAs of plasma. We confirmed 23 miRNAs reported earlier as AD biomarker candidates in blood and found 26 novel differential miRNAs between AD and control subjects. For representative 15 differential miRNAs, the TargetScan, MirTarBase and KEGG database analysis indicated putative protein targets among such AD hallmarks as MAPT (Tau), proteins involved in amyloidogenic proteolysis, and in apoptosis. These 15 miRNAs were verified in separate, subsequent subject groups. Finally, 6 miRNAs (3 not yet reported in AD context and 3 reported in AD blood) were selected as the most promising biomarker candidates differentiating early AD from controls with the highest fold changes (from 1.32 to 14.72), consistent significance, specificities from 0.78 to 1 and sensitivities from 0.75 to 1. (patent pending, PCT/IB2016/052440).

Phantom criteria for qualification of brain FDG and amyloid PET across different cameras

Background

While fluorodeoxyglucose (FDG) and amyloid PET is valuable for patient management, research, and clinical trial of therapeutics on Alzheimer’s disease, the specific details of the PET scanning method including the PET camera model type influence the image quality, which may further affect the interpretation of images and quantitative capabilities. To make multicenter PET data reliable and to establish PET scanning as a universal diagnostic technique and a verified biomarker, we have proposed phantom test procedures and criteria for optimizing image quality across different PET cameras.

Results

As the method, four physical parameters (resolution, gray-white contrast, uniformity, and image noise) were selected as essential to image quality for brain FDG and amyloid PET and were measured with a Hoffman 3D brain phantom and a uniform cylindrical phantom on a total of 12 currently used PET models. The phantom radioactivity and acquisition time were determined based on the standard scanning protocol for each PET drug (FDG, ¹¹C-PiB, ¹⁸F-florbetapir, and ¹⁸F-flutemetamol). Reconstruction parameters were either determined based on the methods adopted in ADNI, J-ADNI, and other research and clinical trials or optimized based on measured phantom image parameters under various reconstruction conditions.

As the result, phantom test criteria were proposed as follows: (i) 8 mm FWHM or better resolution and (ii) gray/white %contrast ≥55 % with the Hoffman 3D brain phantom and (iii) SD of 51 small region of interests (ROIs) ≤0.0249 (equivalent to 5 % variation) for uniformity and (iv) image noise (SD/mean) ≤15 % for a large ROI with the uniform cylindrical phantom. These criteria provided image quality conforming to those multicenter clinical studies and were also achievable with most of the PET cameras that are currently used.

Conclusions

The proposed phantom test criteria facilitate standardization and qualification of brain FDG and amyloid PET images and deserve further evaluation by future multicenter clinical studies.

Radiopharmaceuticals in positron emission tomography: present situation and future perspectives

Positron emission tomography (PET) is an imaging technique that has grown greatly in recent years. PET is considered a fundamental tool in oncology, and it also has indications in other fields such as neurology and cardiology. Although ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) is the radiopharmaceutical most widely used in PET, the availability of new radiotracers has been a key element in the expansion of the use of PET. These new radiopharmaceuticals have made it possible to study different biological targets that are essential for obtaining greater knowledge and better characterization of different diseases and have thus contributed to the research and development of different therapeutic agents. This article provides a description of different PET radiopharmaceutical, structured according to their areas of application. Some of these radiotracers are already commercially available, whereas others are still under research or pending approval by regulatory bodies.

A systematic review of amyloid-β peptides as putative mediators of the association between affective disorders and Alzheimer׳s disease

BACKGROUND

Affective disorders are associated with an increased occurrence of cognitive deficits and have been linked to cognitive impairment and Alzheimer׳s disease. The putative molecular mechanisms involved in these associations are however not clear. The aim of this systematic review was to explore clinically founded evidence for amyloid-β peptides in cerebrospinal fluid and blood as putative biomarkers for affective disorders.

METHOD

Systematic searches in Embase and PubMed databases yielded 23 eligible, observational studies.

RESULTS

Despite inconsistencies that were partly ascribed to the application of different assay formats, study results indicate a potentially altered amyloid-β metabolism in affective disorder.

LIMITATIONS

Since most studies used a cross-sectional design, causality is difficult to establish. Moreover, methodological rigor of included studies varied and several studies were limited by very low sample numbers. Finally, different assays for amyloid-β were utilized in the different studies, thus hampering comparisons.

CONCLUSION

To unravel possible risk relations and causalities between affective disorder and Alzheimer׳s disease and to determine how amyloid-β concentrations change over time and are associated with cognition as well as affective symptomatology, future research should include prospective, longitudinal studies, implemented in large study populations, where peripheral and central amyloid-β ratios are quantified concomitantly and continuously across various affective phases. Also, to enable inter-survey comparisons, the use of standardized pre-analytical/analytical procedures is crucial.

Using neuroimaging to inform clinical practice for the diagnosis and treatment of mild cognitive impairment

Advances in structural and functional neuroimaging techniques have unquestionably improved understanding of the development and progression of Alzheimer disease (AD), with evidence supporting regional (and network) change that underlies cognitive decline across the "healthy" aging/mild cognitive impairment (MCI)/AD spectrum. This review focuses on visual rating scales and volumetric analyses that could be easily integrated into clinical practice, followed by a review of functional neuroimaging findings suggesting that widespread cerebral dysfunction underlies the learning and memory deficits in MCI. Evidence of preserved neuroplasticity in this population and that cognitive rehabilitation techniques may capitalize on this plasticity to improve cognition in those with MCI is also discussed.

Gender, age-related, and regional differences of the magnetization transfer ratio of the cortical and subcortical brain gray matter

PURPOSE

To explore gender, age-related, and regional differences of magnetization transfer ratio (MTR) of brain cortical and subcortical gray matter (GM).

MATERIALS AND METHODS

In all, 102 healthy subjects (51 women and 51 men; range 25-84 years) were examined with 3-mm thick MT images. We assessed MTR in automatically segmented GM structures including frontal, parietal-insular, temporal, and occipital cortex, caudate, pallidus and putamen, and cerebellar cortex. A general linear model analysis was conducted to ascertain the linear and quadratic relationship among the MTR and gender, age, and anatomical structure.

RESULTS

The effect of gender was borderline (P = 0.07) in all GM structures (with higher MTR values in men), whereas age showed a significant linear as well as quadratic effect in all cortical and subcortical GM structures (P ≤ 0.001). Quadratic age-related decrease in MTR began at about 40 years of age. Mean and standard deviation (SD) of MTR had the following decreasing order: thalamus (58.3 + 0.8), pallidus (56.8 ± 1.3), caudate (55.5 ± 1.6) and putamen (54.6 ± 1.1); temporal (56.8 ± 0.9), parietal-insular (56.8 ± 1.1), frontal (56.5 ± 1.1), occipital (55.4 ± 1.0) and cerebellar (53.2 ± 1.0) cortex. In post-hoc testing, all regional pairwise differences were statistically significant except pallidus vs. temporal or parietal-insular cortex, caudate vs. occipital cortex, frontal vs. parietal-insular or temporal cortex.

CONCLUSION

MTR of the cortical and subcortical brain GM structures decreases quadratically after midlife and shows significant regional differences.

Amyloid PET Imaging: MCI and AD

PET with "β-amyloid-specific" molecular imaging probes is proposed for the measurement of brain β-amyloid protein amyloidosis in the new guidelines for diagnosis of Alzheimer disease (AD) at different levels of disease progression. This article discusses limitations of this proposed use pointing to unresolved issues and inconsistencies between PET scan results and correlation with other biomarkers, and with postmortem histopathological studies. These unresolved issues do not warrant the conclusion that PET imaging with "β-amyloid-specific" molecular imaging probes can be used as a biomarker in AD or in the various stages of disease progression.

Molecular neuroimaging in degenerative dementias

In the context of the limitations of structural imaging, brain perfusion and metabolism using SPECT and PET have provided relevant information for the study of cognitive decline. The introduction of the radiotracers for cerebral amyloid imaging has changed the diagnostic strategy regarding Alzheimer's disease, which is currently considered to be a "continuum." According to this new paradigm, the increasing amyloid load would be associated to the preclinical phase and mild cognitive impairment. It has been possible to observe "in vivo" images using 11C-PIB and PET scans. The characteristics of the 11C-PIB image include specific high brain cortical area retention in the positive cases with typical distribution pattern and no retention in the negative cases. This, in combination with 18F-FDG PET, is the basis of molecular neuroimaging as a biomarker. At present, its prognostic value is being evaluated in longitudinal studies. 11C-PIB-PET has become the reference radiotracer to evaluate the presence of cerebral amyloid. However, its availability is limited due to the need for a nearby cyclotron. Therefore, 18F labeled radiotracers are being introduced. Our experience in the last two years with 11C-PIB, first in the research phase and then as being clinically applied, has shown the utility of the technique in the clinical field, either alone or in combination with FDG. Thus, amyloid image is a useful tool for the differential diagnosis of dementia and it is a potentially useful method for early diagnosis and evaluation of future treatments.

Meta-analysis of amyloid-cognition relations in cognitively normal older adults

OBJECTIVE

We conducted a meta-analysis of relationships between amyloid burden and cognition in cognitively normal, older adult humans.

METHODS

Methods of assessing amyloid burden included were CSF or plasma assays, histopathology, and PET ligands. Cognitive domains examined were episodic memory, executive function, working memory, processing speed, visuospatial function, semantic memory, and global cognition. Sixty-four studies representing 7,140 subjects met selection criteria, with 3,495 subjects from 34 studies representing independent cohorts. Weighted effect sizes were obtained for each study. Primary analyses were conducted limiting to independent cohort studies using only the most common assessment method (Pittsburgh compound B). Exploratory analyses included all assessment methods.

RESULTS

Episodic memory (r = 0.12) had a significant relationship to amyloid burden. Executive function and global cognition did not have significant relationships to amyloid in the primary analysis of Pittsburgh compound B (r = 0.05 and r = 0.08, respectively), but did when including all assessment methods (r = 0.08 and r = 0.09, respectively). The domains of working memory, processing speed, visuospatial function, and semantic memory did not have significant relationships to amyloid. Differences in the method of amyloid assessment, study design (longitudinal vs cross-sectional), or inclusion of control variables (age, etc.) had little influence.

CONCLUSIONS

Based on this meta-analytic survey of the literature, increased amyloid burden has small but nontrivial associations with specific domains of cognitive performance in individuals who are currently cognitively normal. These associations may be useful for identifying preclinical Alzheimer disease or developing clinical outcome measures.

Preclinical characterization of a novel class of 18F-labeled PET tracers for amyloid-β

Imaging of amyloid-β (Aβ) plaques by PET is more and more integrated into concepts for Alzheimer disease (AD) diagnosis and drug development. The objective of this study was to find novel chemical entities that can be transformed into (18)F-labeled Aβ tracers with favorable brain washout kinetics and low background signal.

METHODS

High-throughput screening of a large chemical library was used to identify new ligands for fibrillar aggregates of Aβ(1-42) peptide. Thirty-two fluorinated derivatives were synthesized and tested for their affinity toward AD brain homogenate. Twelve ligands have been radiolabeled with (18)F. The pharmacokinetic properties of the radioligands were investigated in mouse and monkey biodistribution studies. Binding characteristics were determined by autoradiography of AD brain sections in vitro and using amyloid precursor protein transgenic mice in vivo.

RESULTS

The systematic search for Aβ imaging agents revealed several fluorinated derivatives with nanomolar affinity for Aβ. The fluoropyridyl derivative BAY 1008472 showed a high initial brain uptake (6.45 percentage injected dose per gram at 2 min) and rapid brain washout (ratio of percentage of injected dose per gram of tissue at 2 and 30 min after injection, 9.2) in mice. PET studies of healthy rhesus monkeys confirmed the high initial brain uptake of BAY 1008472 (2.52 standardized uptake value at peak) and a fast elimination of total radioactivity from gray and white matter areas (ratio of standardized uptake value at peak uptake and 60 min 11.0). In autoradiographic analysis, BAY 1008472 selectively detected Aβ deposits in human AD brain sections with high contrast and did not bind to τ- or α-synuclein pathologies. Finally, ex vivo autoradiography of brain sections from amyloid precursor protein-transgenic mice confirmed that BAY 1008472 is indeed suitable for the in vivo detection of Aβ plaques.

CONCLUSION

A new chemical class of Aβ tracers has been identified by high-throughput screening. The fluoropyridyl derivative BAY 1008472 shows a favorable preclinical profile including low background binding in gray and white matter. These properties might qualify this new tracer, in particular, to detect subtle amounts or changes of Aβ burden in presymptomatic AD and during therapy.