Search

Background: a‐synuclein aggregation is a pathologic hallmark of several Lewy body disorders with differing symptoms. This symptom heterogeneity may be due to varying underlying disease trajectories. Here, we use a data‐driven approach to identify and characterize distinct spatial‐temporal progression patterns of α‐synuclein deposition. Method: The Subtype and Stage Inference (SuStaIn) model evaluates the optimal grouping of individuals into subtypes with distinct orderings of events/stages. We applied Ordinal SuStaIn to post‐mortem Lewy‐type α‐synuclein (LTS) density scores measured in 10 regions (including olfactory bulb and tract [OBT], limbic, brainstem, and cortical regions) of 814 subjects from the Sun Health Research Institute Brain Donation Program (Table‐1). The optimal SuStaIn model was selected according to the Cross‐Validation Information Criteria. We selected 781 subjects with stage>0 (nexcluded = 13) and subtype probability>50% (nexcluded = 20) to assess differences in subtypes. Hierarchical multinomial logistic regressions were performed to compare subtypes on (1) age at death, sex, APOE‐e4 copies, and post‐mortem interval; (2) total amyloid plaque load; and (3) total tau tangle load. In a subset of individuals (n = 355), we additionally compared subtypes on MMSE scores and the motor section of the Unified Parkinson Disease Rating Scale (UPDRS). All models were adjusted for SuStaIn stage and significant predictors of preceding models. Result: SuStaIn identified three subtypes with distinct LTS trajectories, which we termed "OBT/Limbic‐first", "OBT/Brainstem‐first", and "OBT‐sparing/Brainstem‐first" based on the first regions to show deposition (Figure‐1). The majority of individuals were assigned to OBT/Limbic‐first (N = 475,60.8%), followed by OBT/Brainstem‐first (N = 165,21.1%), and OBT‐sparing/Brainstem‐first (N = 141,18.1%) subtypes (Table‐1). We observed a higher frequency of homozygous APOE‐e4 carriers and amyloid and tau burden in the OBT/Limbic‐first subtype compared to the OBT/Brainstem‐first (ßAPOE‐e4 = 0.82±0.40, pAPOE‐e4 =.042; ßamyloid= 0.08±0.42, pamyloid<.001; ßtau= 0.17±0.03, ptau<.001) and OBT‐sparing/Brainstem‐first (ßAPOE‐e4 = 0.98±0.48, pAPOE‐e4 =.042; ßamyloid= 0.14±0.27, pamyloid<.001; ßtau= 0.18±0.03, ptau<.001) subtypes. In addition, the OBT/Brainstem‐first subtype showed a higher amyloid burden (ß = 0.06±0.02, p =.012) and higher MMSE (ß = 0.08±0.04, p = 0.030) compared to the OBT‐sparing/Brainstem‐first (Table‐1, Figure‐2). Conclusion: In this large post‐mortem SuStaIn study, we identified three subtypes of α‐synuclein deposition, possibly reflecting different epicenters and spreading trajectories of misfolded α‐synuclein, where OBT/Limbic‐first was associated with Alzheimer's co‐pathology. Follow‐up work will study how these subtypes differ in clinical diagnosis and symptomatology. [ABSTRACT FROM AUTHOR]

Background: Two main complexities of Alzheimer's Disease (AD) arise from pathological changes (such as amyloid accumulation) occurring a decade or more before symptom onset, and AD being a highly heterogeneous disease with various clinical subtypes. Subtype and Stage Inference (SuStaIn), a data‐driven modelling technique, has been previously applied to ADNI to better understand AD. Here we apply SuStaIn to Insight46, a sub‐study of the British 1946 Birth Cohort to detect different trajectories of preclinical neurodegeneration. Method: SuStaIn was applied to cross‐sectional MRI, amyloid PET, plasma biomarkers, and cognition (Preclinical Alzheimer Cognitive Composite [PACC]) from 466 study members (Table 1). 376 Amyloid‐negative individuals (computed as Lane et al. 2019) were designated as controls. SuStaIn subtypes were also estimated in ADNI (N=789; Tables 1 and 2) for comparison. T‐tests and Chi‐squared tests were used to assess subtype differences in baseline APOE, amyloid status, white matter hyperintensity volume (WMHV), and rates of change in whole brain, ventricular and hippocampal volume calculated using the boundary shift integral. Result: Figure 1 shows three subtypes uncovered by SuStaIn: 'Typical', 'Subcortical' and 'Cortical'. The 'Typical' subtype appears to reflect a typical AD progression pattern, with early involvement of Aβ40/42 ratio and Neurofilament light (NFL), followed by the PACC and finally, plasma tau. MRI volumetric changes appear later. Both 'Subcortical' and 'Cortical' subtypes are characterised by earlier abnormality in MRI volumes. Figure 2 shows that subjects assigned to the 'Cortical' subtype had significantly higher WMHV than the other two subtypes (p‐value < 0.002). No other significant differences in baseline characteristics were found. Figure 3 shows 'Typical' subtype has significantly higher rates of ventricular expansion and whole brain atrophy compared with 'Subcortical' subtype (p‐values = 0.001). Conclusion: This study reveals three different subtypes ('Typical', 'Subcortical' and 'Cortical') of presumed neurodegeneration in an elderly, preclinical cohort; these are consistent with those found in AD patient cohorts, e.g. ADNI, meaning that these methods may have utility in population based cohorts. These results suggest that SuStaIn may be helpful in determining different subtypes of neurodegeneration at a very early stage with implications for recruitment to, and powering of, clinical trials. [ABSTRACT FROM AUTHOR]

Background: Alzheimer's disease is heterogeneous in clinical syndrome, biomarker signature, and pathology. Disentangling this heterogeneity is challenging because multiple underlying factors contribute to the observed heterogeneity, including variability in disease subtype and disease stage. Method: Subtype and Stage Inference (SuStaIn) is an unsupervised learning technique that combines clustering and disease progression modelling to identify subgroups of individuals (subtypes) with distinct progression patterns (stages) from cross‐sectional datasets. We describe the SuStaIn algorithm and recent enhancements and review previous studies using SuStaIn to analyse Alzheimer's heterogeneity across a range of measures, including lobar and subcortical structural MRI volumes, tau PET, white matter hyperintensities, CSF and cognitive scores. We present new data applying SuStaIn to a more fine‐grained set of structural MRI features than has been used previously, extracting volumes from 34 cortical regions. 631 amyloid‐positive individuals based on PET or CSF at their most recent visit were selected from the ADNI dataset. 153 amyloid‐negative cognitively normal individuals were used as a control population. SuStaIn was used to estimate subtypes with distinct progression patterns, determine the optimal number of subtypes, and to subtype and stage individuals. Result: Across previous studies SuStaIn identified three‐four subtypes depending on the choice of biomarkers. Across the 34 cortical regions included in this study, SuStaIn identified a hierarchical organisation of subtypes (Figure 1), with the optimal solution consisting of four subtypes (Figure 2): a 'typical' subtype (N=116), a 'temporo‐parietal subtype' (N=100), a 'fronto‐temporo‐parietal' subtype (N=66) and a 'frontal' subtype (N=19). We observed differences between the subtypes in age and number of APOE4 alleles (Table 1). Individuals assigned to the typical and frontal subtypes were older (p<0.001 typical; p<0.05 frontal), and individuals assigned to the temporo‐parietal and fronto‐temporo‐parietal subtypes were younger (p<0.001 temporo‐parietal; p<0.001 fronto‐temporo‐parietal). The temporo‐parietal subtype had the greatest number of APOE4 alleles (p<0.001), whilst the fronto‐temporo‐parietal subtype had the least number (p<0.05), potentially reflecting an atypical Alzheimer's subtype. The subtype and stage assignments were longitudinally consistent (Figure 3). Conclusion: Our results highlight the substantial heterogeneity of Alzheimer's disease and demonstrate the utility of SuStaIn for identifying Alzheimer's subgroups and their relationship with demographics, genetics, and cognition. [ABSTRACT FROM AUTHOR]

Background: Pathological aggregation of tar DNA‐binding protein 43 (TDP‐43) in the brain is the primary cause of many cases of frontotemporal lobar degeneration (FTLD), amyotrophic lateral sclerosis (ALS) and limbic‐predominant age‐related TDP‐43 encephalopathy (LATE). It is therefore imperative to establish empirical staging systems to characterize and distinguish stereotypical patterns and commonplace deviations of different TDP‐43 proteinopathies. Method: We use ordinal ratings of TDP‐43 burden from 19 brain regions to perform data‐driven disease progression modeling (SuStaIn) to find the most likely trajectories for FTLD‐TDP (n = 108), ALS (n = 137) and LATE (n = 283) from the CNDR Brain Bank at the University of Pennsylvania. Subtype number was defined using cross‐validated information criterion. Each individual was assigned a subtype and stage. Multivariate OLS models tested differences between subtypes. Stages were compared to age and existing staging schemes. Cross‐validated logistic regression was used for 3‐way classification using SuStaIn information only. Result: SuStaIn provided data‐driven staging of TDP‐43 proteinopathies complementing previously described human‐defined staging schema, further providing additional detail (Fig1A‐C; Fig3A‐C). SuStaIn also identified two distinct subtypes within FTLD‐TDP and a further two within ALS (Fig1D). FTLD‐TDP subtypes differed in TDP‐43 type and Alzheimer's disease pathology (Table1); ALS subtypes were differentiated by age (Table 2) and by antemortem clinical characteristics. No subtypes were observed for the LATE group. Progression along data‐driven stages was positively associated with age in LATE individuals, but negatively associated with age in individuals with FTLD‐TDP (Fig2). Using only regional TDP‐43 severity, our data driven model could distinguish individuals diagnosed with ALS, FTD or LATE with a cross‐validated balanced precision of 0.93 and balanced recall of 0.92, and these metrics improved to 0.95 and 0.96 when combined with a logistic regression model (Fig3). Very little stage overlap was found between FTLD‐TDP and LATE, but stages that did overlap showed subtly different patterns (Fig4) Conclusion: We provide an empirical pathological staging system for ALS, FTLD‐TDP and LATE, which is sufficient for staging and accurate classification. We demonstrate that there is substantial heterogeneity amongst ALS and FTLD‐TDP progression patterns, whilst LATE exhibits a homogeneous progression pattern. [ABSTRACT FROM AUTHOR]

Background: The progression of Alzheimer's Disease (AD) is heterogeneous with different brain regions manifesting pathology at different times during the disease. Differential sequences of grey matter (GM) volume changes, defining imaging‐based subtypes, may represent alternative disease origins, pathways or severity. Additionally, volumes of white matter hyperintensities (WMH), which increase with age, are associated with AD. In this study we used a machine learning method to estimate three types of AD trajectories and assessed their WMH load and longitudinal prognosis. Method: Subtype and Stage Inference (SuStaIn; Young et al., https://doi.org/10.1038/s41467‐018‐05892‐0) was used to estimate data‐driven disease progresion subtypes from cross‐sectional GM volumes in 795 subjects (from ADNI1/GO/2 with baseline 3T MRI scans processed with FreeSurfer 5.1 passing data quality control). Following additional quality control WMH volumes for 575 of these ADNI subjects were calculated using a Bayesian Model Selection method BaMoS (Sudre et al., https://doi.org/10.1109/TMI.2015.2419072) from three diagnostic groups: Healthy Controls (HC; n=181), Mild Cognitive Impairment (MCI; n=310) and Alzheimer's Disease (AD; n=85). The association between SuStaIn subtypes and WMH and diagnostic conversion rates were evaluated using multiple pairwise comparisons and survival analysis, respectively. Result: SuStaIn identified three subtypes with distinct atrophy patterns: typical (atrophy begins in hippocampus and amygdala), cortical (atrophy starts in the temporal lobe, followed by cingulate and insula) and subcortical (atrophy begins in pallidum, putamen and caudate) (Figure 1). Participants with the subcortical subtype exhibited a significantly slower rate of clinical progression to MCI or AD (Figure 2, Table 1). Furthemore, subjects with the subcortical subtype showed lower median WMH (Figure 3a), a pattern observed across the diagnostic groups (Figure 3b). Conclusion: Participants classified as the subcortical subtype were least likely to have white matter structural changes associated with neurodegeneration and displayed the slowest rate of clinical conversion. While late hippocampal involvement in the subcortical subtype may explain less severe clinical symptoms and slower disease progression, the reason for lower WMH associated with this subtype trajectory is unknown. Further studies are needed to investigate the link between WMH and AD atrophy subtypes. [ABSTRACT FROM AUTHOR]

Background: The preclinical phase of Alzheimer's disease (AD), where pathology slowly accumulates years before cognitive impairment becomes apparent, could offer a treatment window with the greatest potential to preserve cognitive function before downstream pathological processes gather momentum. Characterizing when biomarker trajectories deviate from normal ageing, and the heterogeneity therein, could facilitate targeted trial recruitment and improved biomarker‐based evidence of disease modification. However, reliably identifying early abnormal changes can be challenging due to various confounds, such as age and vascular factors, as well as disease heterogeneity. Method: Data from the following cohorts were analysed: (1) individuals at risk for, or affected by, autosomal dominant AD (ADAD), including the Dominantly Inherited Alzheimer Network; (2) Insight 46, a neuroimaging substudy of the MRC National Survey of Heath and Development; (3) pre‐randomization study data from the Anti‐Amyloid Treatment in Asymptomatic Alzheimer's Disease (A4); and (4) the Alzheimer's Disease Neuroimaging Initiative (ADNI). Linear regression assessed differences between at‐risk groups (biomarker‐ or mutation‐positive) and normal ageing. Longitudinal changepoint models estimated when atrophy rates deviate from normal ageing trajectories. Normative modelling, based on large reference datasets, identified outliers of regional atrophy assessed within‐group heterogeneity at the individual level. Data‐driven disease progression models (DPMs) were used to estimate quantitative signatures of AD, including imaging biomarkers. Result: DPMs reveal that amyloid markers deviate first in ADAD, roughly ten years before neurodegenerative markers. Changepoint models indicate that atrophy rates in ADAD become abnormal 3‐8 years before expected symptom onset, with longitudinal tau PET data suggesting a similar trend. While cross‐sectional volumetric measures showed no evidence of association with amyloid markers, atrophy rates were independently related to both amyloid positivity and markers of cerebrovascular disease, indicating similar, additive effects. Both DPMs and normative modelling found evidence of heterogeneity within imaging biomarkers in ADNI and A4, which explained variability in cognitive decline observed in at‐risk participants. Conclusion: Biomarker data from preclinical AD suggests a long temporal gap between amyloidosis and subsequent changes. This represents a treatment opportunity to remove amyloid while minimising irreversible damage. There is evidence of heterogeneity within the biomarker trajectories, which could impact the ability to detect disease modification. [ABSTRACT FROM AUTHOR]

Background: The Braak staging scheme describes a stereotypical spread of tau pathology in Alzheimer's disease (AD). However, apparent subtypes described at autopsy and clinical variants of AD both suggest variability in the pattern of tau spreading across the population. To address this discrepancy, we applied a validated spatiotemporal clustering algorithm to a large, multisite dataset of tau‐PET images. We verified these findings both longitudinally and on a separate tau‐PET dataset using a different radiotracer, and we characterized the phenotype and potential biological determinants of each subtype. Method: Subtype and Stage Inference (SuStaIn, Young et al., 2018) is a data‐driven, probabilistic algorithm that combines disease progression modeling with clustering to extract distinct spatiotemporal patterns (subtypes) from large cross‐sectional datasets. We applied SuStaIn to a multisite dataset (n=1143) of [18F]‐flortaucipir tau‐PET baseline scans to identify spatiotemporal subtypes of tau spreading. We investigated differences in demographic, cognitive and genetic measures. In a subsample of individuals with multiple tau‐PET scans over time (n=536), we tested longitudinal subtype stability and change in disease stage over time. We verified these subtypes in a second single‐site dataset (n=509) of [18F]‐RO948 tau‐PET scans. Finally we used the Epidemic Spreading Model (ESM, Iturria‐Medina et al., 2014) and imaging or transcriptomic analysis to determine if anatomical brain networks or specific cell‐types, respectively, could explain the differing spatial patterns across subtypes. Result: In those subjects with significant tau‐PET signal in the brain, SuStaIn identified four spatiotemporal subtypes of tau‐PET spreading: Typical (worse memory, more APOE4 carriers), limbic‐sparing (younger, less APOE4 carriers), posterior (worse visuospatial cognition), and lateral‐temporal (worse language and executive function scores; Fig 1A, Fig 2). These same four subtypes were observed in a separate cohort using a different radiotracer (similarity 0.7–0.9; Fig 1B). 86% of subjects exhibited the same subtype at multiple visits (Fig 1C), and of that group, 83% advanced in disease stage or remained stable within model uncertainty. The ESM suggested different epicenters for each subtype, and each subtype pattern exhibited a different profile of cell‐type enrichment (Fig 3). Conclusion: Four stable tau‐spreading phenotypes were observed with distinct cognitive profiles, which may be characterized by vulnerability of different corticolimbic networks involving distinct cell classes. [ABSTRACT FROM AUTHOR]

Background: Several studies contend that some Alzheimer's disease (AD) subjects develop early‐stage tau pathology before amyloid pathology1,2, supporting a hypothesis that there are multiple distinct subtypes of amyloid and tau pathology progression within AD. We investigated this hypothesis, applying data‐driven disease subtyping models to both amyloid and tau PET as well as CSF measures. Method: We performed two separate analyses using cross‐sectional data from ADNI. The first was a tau‐PET‐based analysis, which used eight regional amyloid PET (AV‐45) SUVRs and ten tau PET (AV‐1451) SUVRs from 402 subjects with both scans at the same visit. The second was a tau‐CSF‐based analysis, substituting CSF‐based prediction of tau PET SUVR (see Figure 2) in place of actual tau PET to give a larger dataset of 996 subjects. We used the Subtype and Stage Inference (SuStaIn3) algorithm to infer disease progression subtypes and individuals' disease stages, and characterized demographic, cognitive and CSF differences across subtypes. Result: We found two PET‐based subtypes in the tau‐PET‐based analysis: an amyloid‐first (84% of subjects) subtype and a tau‐first (16% of subjects) subtype in which Braak stage I‐III related tau SUVRs (hippocampus, amygdala, entorhinal cortex) become abnormal first (Figure 1A,B). The tau‐CSF‐based analysis confirms these subtypes: amyloid‐first (83% of subjects) and tau‐first (17% of subjects; Figure 2A,B). There were no significant differences in demographics or subject stages between subtypes in the tau‐PET‐based analysis (Figure 1C,D). In the tau‐CSF‐based analysis there was a small difference in age (amyloid‐first subjects 2.4 years older, Cohen's f2 = 0.02, p < 0.05). Amyloid‐first individuals also had slightly worse executive function (Cohen's f2 = 0.02, p < 0.05; Figure 3C) and, as expected, more abnormal CSF amyloid (Cohen's f2 = 0.10, p < 0.001; Figure 3D) while tau‐first individuals had more abnormal CSF tau (Cohen's f2 = 0.06, p < 0.001; Figure 3D). Conclusion: We identified amyloid‐first and tau‐first AD subtypes consistently across PET and CSF biomarkers. Our data‐driven approach supports the existence of a subtype of AD with tau accumulation prior to amyloid pathology. References: 1. Duyckaerts, et al., 2015, Acta Neuropath; 2. Weigand, et al., 2019, Brain Comm; 3. Young, et al., 2018, Nature Comm. [ABSTRACT FROM AUTHOR]

Background: Mutations in the MAPT gene are known to cause frontotemporal dementia (FTD), but there is heterogeneity in FTD phenotype across individuals. Here we used an unsupervised learning algorithm – Subtype and Stage Inference (SuStaIn) – to relate phenotypic heterogeneity to specific mutations in the MAPT gene. Method: SuStaIn evaluates the optimal grouping of individuals into disease subtypes, where each subtype consists of a sequence (set of stages) in which biomarkers transition between different z‐scores. We applied SuStaIn to cross‐sectional regional brain volumes extracted from T1‐weighted MRI data from MAPT carriers in the GENFI study to find the best stratification of the data into subtypes, and the temporal progression of each subtype. We used data from 82 MAPT carriers (57 presymptomatic and 25 symptomatic) to identify subtypes and data from a control group of 300 non‐carriers to derive z‐scores. We subtyped and staged individuals at up to five annual follow‐up visits to assess the consistency of the subtypes longitudinally. We compared the specific mutations and clinical and neuropsychological test scores of individuals assigned to each subtype. Result: SuStaIn identified two groups of MAPT carriers with distinct atrophy patterns (Figure 1), which we termed a 'temporal' subtype and a 'frontotemporal' subtype. The subtype assignments were consistent at follow‐up visits (Table 1): there were no individuals that changed from the temporal to the frontotemporal subtype or vice‐versa. Subtype assignment was strongly associated with IVS10+16, R406W and P301L mutations (Table 2): there was a one‐to‐one mapping between IVS10+16 and R406W mutations and the temporal subtype, and a near one‐to‐one mapping between P301L mutations and the frontotemporal subtype. The temporal subtype was associated with memory problems, whereas the frontotemporal subtype was associated with worse performance on tests of attention and visuospatial skills (Table 3). Conclusion: Our results demonstrate the utility of SuStaIn for identifying disease subgroups and associating imaging patterns with genetics and cognition. We show that different MAPT mutations give rise to distinct atrophy patterns and clinical syndromes, providing insights into the underlying disease biology, and potential utility for patient stratification. [ABSTRACT FROM AUTHOR]

Background: Efficient visualisation of brain structure, function and pathology is crucial for understanding the mechanisms underlying neurodegenerative diseases such as Alzheimer's disease (AD). However, visualisation of dynamic brain processes such as pathology spread in AD is difficult due to the slow progression of pathological cascades at multiple scales and over extended periods of time. Current visualisation software (e.g. Freesurfer or SPM) are excellent for interactive visualisation of single brain scans, but for a study population, do not enable easy generation of coloured brain templates or movies showing the temporal progression of pathology. Method: We present BrainPainter (Fig 1), a recently released brain visualisation software that has two key advantages: (1) it can generate brain templates or movies showing dynamic processes in the brain, e.g. propagation of amyloid or tau pathology and (2) it does not require the input data to be in a specialised format, allowing it to be used in conjunction with any neuroimaging analysis software (e.g. Freesurfer, SPM). Result: We highlight two use cases of BrainPainter for existing neuroimaging studies: (i) Fig. 2 – visualisation of the progression of pathology in Alzheimer's disease and (ii) Fig. 3 – visualisation of pathology in subcortical regions in Huntington's disease. We further present two novel functionalities of BrainPainter: (iii) generation of movies showing progression of atrophy in Alzheimer's disease (e.g. https://brainpainter.csail.mit.edu/#examples) and (iv) Fig. 4 – visualisation of individual regions‐of‐interest based on three supported atlases (Desikan‐Killiany, Destrieux and Trourville). Conclusion: BrainPainter is customisable, easy to use, and runs straight from the web browser: https://brainpainter.csail.mit.edu. It can generate visualisations of imaging data from any modality, which can easily be included as images or videos into neuroimaging studies. Through the design of BrainPainter, we further demonstrate the possibility of turning Blender, the powerful 3D rendering engine, into a useful tool for the neuroscience community. [ABSTRACT FROM AUTHOR]

Background: Studies in cells, animals and humans have each provided evidence that, in an Alzheimer's disease (AD) context, tau may be spreading through the brain transneuronally. This information may be useful for modeling the spread of tau in humans. However, such models are confounded by the substantial heterogeneity in spreading patterns observed in AD. In this talk, I will discuss i) application of an epidemic spreading model (ESM) to tau‐PET data to model the spread of tau through the human connectome; ii) application of a spatiotemporal subtyping algorithm to identify separable tau‐spreading patterns; 3) how accounting for these heterogeneous patterns can improve connectome‐based models. Method: The ESM simulates diffusion of an agent through a system of connected brain regions measured with diffusion tractography. We compare simulations across 312 individuals to observed tau‐PET data. We next use the Subtype and Staging Inference (SuStaIn), an algorithm combining disease progression models with clustering, to identify spatiotemporal subtypes of tau spreading across 1143 individuals. We describe demographic, cognitive and genetic associations with each subtype, and evaluate their stability over time and across different radiotracers. Finally, we apply the ESM separately to each AD subtype to assess whether overall model accuracy is improved by accounting for individual subtype. Result: The best‐fitting ESM used the entorhinal cortex as the model epicenter, and explained 70% of the variance in the overall tau‐PET pattern across subjects. SuStaIn identified four spatiotemporal subtypes with differing tau‐PET patterns and phenotypic profiles: Limbic‐predominant, limbic‐sparing, posterior, and lateral‐temporal. These same four subtypes were observed in a separate cohort using a different radiotracer (similarity 0.7‐0.9). 86% of subjects exhibited the same subtype at follow‐up. The ESM selected a different epicenter for each subtype, and accounting for subtype‐specific variation resulted in a 17% improvement in model performance. Conclusion: Connectome‐based models explain the majority of spatial variation in tau‐PET patterns. However, a one‐size‐fits‐all approach fails to account for heterogeneity of tau‐spreading patterns. Four stable phenotypes were observed, which may be characterized by vulnerability of different cortico‐limbic networks. Accounting for this heterogeneity will be necessary for any predictive modeling of tau spreading moving forward. [ABSTRACT FROM AUTHOR]

Background: Accurate prediction of progression in subjects at risk of Alzheimer's disease is crucial for enrolling the right subjects in clinical trials, at the right time. However, an unbiased comparison of state‐of‐the‐art algorithms for predicting disease onset and progression is currently lacking. Method: We present the findings of The Alzheimer's Disease Prediction Of Longitudinal Evolution (TADPOLE) Challenge, which compared the performance of 92 algorithms (Fig. 1) from 33 international teams at predicting the future trajectory of 219 individuals at risk of Alzheimer's disease (demographics in Fig. 2). Challenge participants were required to make predictions, for each month of a 5‐year future time period, of three key outcomes: clinical diagnosis, ADAS‐Cog13, and total volume of the ventricles. Results: No single submission was best at predicting all three outcomes. For clinical diagnosis and ventricle volume prediction, the best algorithms strongly outperformed simple baselines (Fig. 3). However, for ADAS‐Cog13 no single submitted prediction method was significantly better than random guessing. On a limited cross‐sectional dataset emulating clinical trials, performance of the best algorithms at predicting clinical diagnosis decreased only slightly (+3% error) compared to the full longitudinal dataset (Fig. 4). Two consensus methods — mean and median over all predictions — obtained top scores on almost all tasks (Fig. 3‐4). Better‐than‐average performance at predicting clinical progression was associated with the additional inclusion of features from cerebrospinal fluid (CSF) and diffusion tensor imaging (DTI) (Fig. 5). On the other hand, better performance at ventricle volume prediction was associated with inclusion of summary statistics, such as patient‐specific biomarker trends (Fig. 5). Full results can be found on the website https://tadpole.grand‐challenge.org, while code for submissions is being collated by TADPOLE SHARE: https://tadpole‐share.github.io/. Conclusion: Our work suggests that current prediction algorithms are accurate for biomarkers related to clinical diagnosis and ventricle volume, opening up the possibility of cohort refinement in clinical trials for Alzheimer's disease. [ABSTRACT FROM AUTHOR]

Introduction: This work aims to characterize the sequence in which cognitive deficits appear in two dementia syndromes. Methods: Event‐based modeling estimated fine‐grained sequences of cognitive decline in clinically‐diagnosed posterior cortical atrophy (PCA) (n=94) and typical Alzheimer's disease (tAD) (n=61) at the UCL Dementia Research Centre. Our neuropsychological battery assessed memory, vision, arithmetic, and general cognition. We adapted the event‐based model to handle highly non‐Gaussian data such as cognitive test scores where ceiling/floor effects are common. Results: Experiments revealed differences and similarities in the fine‐grained ordering of cognitive decline in PCA (vision first) and tAD (memory first). Simulation experiments reveal that our new model equals or exceeds performance of the classic event‐based model, especially for highly non‐Gaussian data. Discussion: Our model recovered realistic, phenotypical progression signatures that may be applied in dementia clinical trials for enrichment, and as a data‐driven composite cognitive end‐point. [ABSTRACT FROM AUTHOR]

Background: Data‐driven modelling techniques, such as Subtype and Stage Inference (SuStaIn) (Young, et al., 2018), have been developed to disentangle the temporal and phenotypic heterogeneity of neurodegenerative diseases such as Alzheimer's disease (AD) from widely available cross‐sectional datasets such as ADNI. This study builds on early work with SuStaIn by using multimodal data and missing‐data modelling to give a more detailed picture of the landscape of sporadic AD subtypes. Method: SuStaIn was applied to cross‐sectional cerebrospinal fluid (CSF), magnetic resonance imaging (MRI) and positron emission tomography (PET) metrics, and cognitive scores from ADNI to identify multimodal disease subtypes with distinct progression patterns. The dataset consists of 789 subjects that passed quality control (182 cognitively normal (CN), 86 subjective memory concern (SMC), 241 early mild cognitive impairment (MCI), 163 late MCI and 117 AD). Amyloid‐negative (CSF Abeta < 192pg/ml) subjects were chosen as control subjects. A K‐Nearest Neighbour (KNN) algorithm was used to impute missing values (3% of the data), allowing for subtyping and staging of the full dataset. Result: Figure 1 visualises the subtype progression patterns, with the precise model output being shown in Figure 2. Four different subtypes can be inferred from the data: 1) Typical AD with late tau: First changes are seen in CSF and PET measures of b‐amyloid, followed by cognitive scores impairment and finally, brain atrophy. 2) Typical AD: Changes in b‐amyloid closely followed by an increase in tau, both using CSF and PET based biomarkers. Cognitive scores and subcortical areas are affected next. 3) Cortical AD: Cortical atrophy is followed by subcortical atrophy and then increase in CSF and PET amyloid. 4) Cognitive decline with hippocampal atrophy: Mainly defined by a decrease in cognitive tests performance, followed by atrophy of the hippocampus and the amygdala. Conclusion: This study provides a deeper insight into the heterogeneity of AD, revealing four subtypes with differing trajectories of biomarker abnormalities. Unravelling this heterogeneity will be a critical step in developing personalised medicine, allowing for improved power for clinical trials and better information for patients. [ABSTRACT FROM AUTHOR]

Background: Dementia can negatively impact the well‐being and quality of life of patients and informal caregivers. Digital self‐management tools have the potential to support those living with dementia. The aim of this review was to identify digital tools used in dementia care, investigate how they support users, and highlight pertinent clinical and psychosocial outcomes. Method: Six databases were searched (i.e. CINAHL, EMBASE, MEDLINE, PsycINFO, Web of Science, and the Cochrane Review) using free text and equivalent database‐controlled vocabulary terms. Result: Twenty‐six articles were identified reporting 21 digital tools focused mainly on caregivers ‐ thus caregivers form the focus of this review. The digital tools promoted self‐management and support through five main functions: Supporting psychoeducation and self‐help; improving competency and ability; sustaining and promoting well‐being; encouraging behaviour change and providing motivation; and facilitating contact and communication with healthcare professionals and other users. There was mixed‐to‐limited evidence to support a positive influence of digital tools on mental health, burden, quality of life, and caregiver perceived ability. Conclusion: Digital health technology has the potential to support caregivers, however areas such as mental health, burden, and quality of life may be resistant to some digital interventions. It is important to understand issues around accessibility, uptake, and continued use of digital tools and their potential to be integrated within dementia care pathways and routine clinical care. [ABSTRACT FROM AUTHOR]

Background: Neurodegenerative diseases are highly heterogeneous, consisting of multiple disease subtypes with different spatiotemporal patterns of pathology. These patterns evolve dynamically with disease stage, making it challenging to disentangle disease subtypes in vivo from biomarker data where precise staging information is unavailable. Method: Subtype and Stage Inference (SuStaIn) is an unsupervised learning technique that disentangles biomarker heterogeneity into subgroups of individuals (subtypes) with distinct progression patterns (stages). SuStaIn evaluates the optimal grouping of individuals into disease subtypes, where each subtype consists of a sequence in which biomarkers transition between different z‐scores. We used SuStaIn in a number of conditions to identify subgroups of individuals with distinct patterns of brain volume loss using structural MRI data. We present results from several datasets, reviewing results in Alzheimer's disease (ADNI dataset), and presenting new results in genetic frontotemporal dementia (GENFI dataset), and in the aging population (UK Biobank). In each dataset we explored genetic, neuropsychological, clinical and biomarker associations with each of the subtypes identified by SuStaIn. We used longitudinal data to assess the consistency of the subtypes and stages assigned by the SuStaIn model at follow‐up visits. Result: In Alzheimer's disease we identified multiple spatiotemporal patterns of neurodegeneration associated with differences in memory and executive function scores. We found a genetic link between a 'limbic‐predominant' atrophy pattern and type 2 diabetes through a GWAS and analysis of polygenic risk scores. In genetic frontotemporal dementia we identified and confirmed two distinct atrophy patterns amongst C9orf72 mutation carriers (N=182). In MAPT mutation carriers (N=82), we identified two atrophy subtypes that were strongly associated with different MAPT mutations: a 'temporal' subtype, which had a one‐to‐one mapping with IVS10+16 and R406W mutations, and a 'frontotemporal' subtype, which had a near one‐to‐one mapping with P301L mutations. The subtype assignments showed strong stability at follow‐up visits (99% consistency). In the UK Biobank (N=21390), we identified multiple patterns of brain volume loss associated with distinct characteristics. Conclusion: Our results demonstrate the utility of SuStaIn for identifying disease subgroups and associating imaging patterns with genetics and cognition. We show that SuStaIn can provide enhanced patient stratification capability across multiple conditions. [ABSTRACT FROM AUTHOR]