Alexandre Routier

Standardized evaluation of algorithms for computer-aided diagnosis of dementia based on structural MRI: The CADDementia challenge

Algorithms for computer-aided diagnosis of dementia based on structural MRI have demonstrated high performance in the literature, but are difficult to compare as different data sets and methodology were used for evaluation. In addition, it is unclear how the algorithms would perform on previously unseen data, and thus, how they would perform in clinical practice when there is no real opportunity to adapt the algorithm to the data at hand. To address these comparability, generalizability and clinical applicability issues, we organized a grand challenge that aimed to objectively compare algorithms based on a clinically representative multi-center data set. Using clinical practice as the starting point, the goal was to reproduce the clinical diagnosis. Therefore, we evaluated algorithms for multi-class classification of three diagnostic groups: patients with probable Alzheimers disease, patients with mild cognitive impairment and healthy controls. The diagnosis based on clinical criteria was used as reference standard, as it was the best available reference despite its known limitations. For evaluation, a previously unseen test set was used consisting of 354 T1-weighted MRI scans with the diagnoses blinded. Fifteen research teams participated with a total of 29 algorithms. The algorithms were trained on a small training set (n=30) and optionally on data from other sources (e.g., the Alzheimers Disease Neuroimaging Initiative, the Australian Imaging Biomarkers and Lifestyle flagship study of aging). The best performing algorithm yielded an accuracy of 63.0% and an area under the receiver-operating-characteristic curve (AUC) of 78.8%. In general, the best performances were achieved using feature extraction based on voxel-based morphometry or a combination of features that included volume, cortical thickness, shape and intensity. The challenge is open for new submissions via the web-based framework: http://caddementia.grand-challenge.org.

Lateral Temporal Lobe: An Early Imaging Marker of the Presymptomatic GRN Disease?

Abstract The preclinical stage of frontotemporal lobar degeneration (FTLD) is not well characterized. We conducted a brain metabolism (FDG-PET) and structural (cortical thickness) study to detect early changes in asymptomatic GRN mutation carriers (aGRN+) that were evaluated longitudinally over a 20-month period. At baseline, a left lateral temporal lobe hypometabolism was present in aGRN+ without any structural changes. Importantly, this is the first longitudinal study and, across time, the metabolism more rapidly decreased in aGRN+ in lateral temporal and frontal regions. The main structural change observed in the longitudinal study was a reduction of cortical thickness in the left lateral temporal lobe in carriers. A limit of this study is the relatively small sample (n = 16); nevertheless, it provides important results. First, it evidences that the pathological processes develop a long time before clinical onset, and that early neuroimaging changes might be detected approximately 20 years before the clinical onset of disease. Second, it suggests that metabolic changes are detectable before structural modifications and cognitive deficits. Third, both the baseline and longitudinal studies provide converging results implicating lateral temporal lobe as early involved in GRN disease. Finally, our study demonstrates that structural and metabolic changes could represent possible biomarkers to monitor the progression of disease in the presymptomatic stage toward clinical onset.

A Mixed-Effects Model with Time Reparametrization for Longitudinal Univariate Manifold-Valued Data.

Mixed-effects models provide a rich theoretical framework for the analysis of longitudinal data. However, when used to analyze or predict the progression of a neurodegenerative disease such as Alzheimers disease, these models usually do not take into account the fact that subjects may be at different stages of disease progression and the interpretation of the model may depend on some implicit reference time. In this paper, we propose a generative statistical model for longitudinal data, described in a univariate Riemannian manifold setting, which estimates an average disease progression model, subject-specific time shifts and acceleration factors. The time shifts account for variability in age at disease-onset time. The acceleration factors account for variability in speed of disease progression. For a given individual, the estimated time shift and acceleration factor define an affine reparametrization of the average disease progression model. This statistical model has been used to analyze neuropsychological assessments scores and cortical thickness measurements from the Alzheimers Disease Neuroimaging Initiative database. The numerical results showed that we can distinguish between slow versus fast progressing and early versus late-onset individuals.

Amyloidosis and neurodegeneration result in distinct structural connectivity patterns in mild cognitive impairment

Alzheimers disease (AD) is increasingly considered as a disconnection syndrome. Previous studies of the structural connectome in early AD stages have focused on mild cognitive impaired subjects (MCI), considering them as a homogeneous group. We studied 168 subjects from the Alzheimers Disease Neuroimaging Initiative database (116 MCI and 52 cognitively normal subjects). Biomarker-based stratification using amyloid biomarkers (AV45 PET) and neurodegeneration biomarkers (MRI and FDG PET) led to 4 subgroups based on amyloid positivity (A+/-) and neurodegeneration positivity (N+/-): A-N-, A+N-, A-N+, and A+N+. Using diffusion MRI, we showed that both MCI A-N+ and MCI A+N+ subjects displayed an alteration of the white matter in the fornix and a significant bihemispheric network of decreased connections. These network alterations in MCI A+N+ are stronger and more focal than those of MCI A-N+. Only MCI A+N+ subjects exhibited specific changes in hippocampal connectivity and an AD-like alteration pattern. Our results indicate that the connectome disintegration pattern of MCI subgroups differ with respect to brain amyloid and neurodegeneration. Each of these 2 AD biomarkers induces a connectome alteration that is maximal when they coexist.

Improving efficiency of data assimilation procedure for a biomechanical heart model by representing surfaces as currents

We adapt the formalism of currents to compare data surfaces and surfaces of a mechanical model and we use this discrepancy measure to feed a data assimilation procedure. We apply our methodology to perform parameter estimation in a biomechanical model of the heart using synthetic observations of the endo- and epicardium surfaces of an infarcted left ventricle. We compare this formalism with a more classical signed distance operator between surfaces and we numerically show that we have improved the efficiency of our estimation justifying the use of state-of-the-art computational geometry formalism in the data assimilation measurements processing.

Early cognitive, structural and microstructural changes in c9orf72 presymptomatic carriers before 40 years of age

Importance Presymptomatic carriers of chromosome 9 open reading frame 72 (C9orf72) mutation, the most frequent genetic cause of frontotemporal lobar degeneration and amyotrophic lateral sclerosis, represent the optimal target population for the development of disease-modifying drugs. Preclinical biomarkers are needed to monitor the effect of therapeutic interventions in this population. Objectives To assess the occurrence of cognitive, structural, and microstructural changes in presymptomatic C9orf72 carriers. Design, Setting, and Participants The PREV-DEMALS study is a prospective, multicenter, observational study of first-degree relatives of individuals carrying the C9orf72 mutation. Eighty-four participants entered the study between October 2015 and April 2017; 80 (95%) were included in cross-sectional analyses of baseline data. All participants underwent neuropsychological testing and magnetic resonance imaging; 63 (79%) underwent diffusion tensor magnetic resonance imaging. Gray matter volumes and diffusion tensor imaging metrics were calculated within regions of interest. Anatomical and microstructural differences between individuals who carried the C9orf72 mutation (C9+) and those who did not carry the C9orf72 mutation (C9−) were assessed using linear mixed-effects models. Data were analyzed from October 2015 to April 2017. Main Outcomes and Measures Differences in neuropsychological scores, gray matter volume, and white matter integrity between C9+ and C9− individuals. Results Of the 80 included participants, there were 41 C9+ individuals (24 [59%] female; mean [SD] age, 39.8 [11.1] years) and 39 C9− individuals (24 [62%] female; mean [SD] age, 45.2 [13.9] years). Compared with C9− individuals, C9+ individuals had lower mean (SD) praxis scores (163.4 [6.1] vs 165.3 [5.9]; P = .01) and intransitive gesture scores (34.9 [1.6] vs 35.7 [1.5]; P = .004), atrophy in 8 cortical regions of interest and in the right thalamus, and white matter alterations in 8 tracts. When restricting the analyses to participants younger than 40 years, compared with C9− individuals, C9+ individuals had lower praxis scores and intransitive gesture scores, atrophy in 4 cortical regions of interest and in the right thalamus, and white matter alterations in 2 tracts. Conclusions and Relevance Cognitive, structural, and microstructural alterations are detectable in young C9+ individuals. Early and subtle praxis alterations, underpinned by focal atrophy of the left supramarginal gyrus, may represent an early and nonevolving phenotype related to neurodevelopmental effects of C9orf72 mutation. White matter alterations reflect the future phenotype of frontotemporal lobar degeneration/amyotrophic lateral sclerosis, while atrophy appears more diffuse. Our results contribute to a better understanding of the preclinical phase of C9orf72 disease and of the respective contribution of magnetic resonance biomarkers. Trial Registration clinicaltrials.gov Identifier: NCT02590276

Joint Morphometry of Fiber Tracts and Gray Matter Structures Using Double Diffeomorphisms

This work proposes an atlas construction method to jointly analyse the relative position and shape of fiber tracts and gray matter structures. It is based on a double diffeomorphism which is a composition of two diffeomorphisms. The first diffeomorphism acts only on the white matter keeping fixed the gray matter of the atlas. The resulting white matter, together with the gray matter, are then deformed by the second diffeomorphism. The two diffeomorphisms are related and jointly optimised. In this way, the, first diffeomorphisms explain the variability in structural connectivity within the population, namely both changes in the connected areas of the gray matter and in the geometry of the pathway of the tracts. The second diffeomorphisms put into correspondence the homologous anatomical structures across subjects. Fiber bundles are approximated with weighted prototypes using the metric of weighted currents. The atlas, the covariance matrix of deformation parameters and the noise variance of each structure are automatically estimated using a Bayesian approach. This method is applied to patients with Tourette syndrome and controls showing a variability in the structural connectivity of the left cortico-putamen circuit.

The structure of the mental lexicon: What primary progressive aphasias reveal

&NA; Like recursive syntax, a structured mental lexicon is specific to the human species but its internal organization remains unclear. It is thought to contain information about the semantic, syntactic (e.g., gender) and formal (orthographic/phonological) features of a word. Previous studies suggested that these three components might be separated at the behavioral level and that they might be implemented by temporal cortices. However, the available investigations are based on case reports or small‐cohort studies with patients demonstrating post‐stroke aphasia, and they did not contrast the three lexical components in a directly comparable way. Similarly, functional imaging studies with healthy adults did not compare the lexical components but explored them separately using various tasks. Here we assessed the three components with comparable tasks in a relatively large cohort of 20 patients with primary progressive aphasia (PPA), namely logopenic and semantic PPA, which have been shown to affect the temporal cortex. The same tasks were also applied to 23 healthy adults. We thereby primarily aimed at showing multiple intra‐lexical dissociations at the behavioral level to demonstrate the existence of a threefold segregation within the mental lexicon. We also sought to confirm the temporal‐cortical involvement in the implementation of the lexical components and to characterize differential lexical breakdown in PPA. Lexical components were explored with three implicit processing tasks (semantic, syntactic‐gender, word‐form priming) and with three explicit matching tasks (semantic, syntactic‐gender, word‐form). Our results indicate that the three components are functionally segregated as evidenced by multiple dissociations at the group level, and the individual level, thus substantiating the existence of a threefold structure of the mental lexicon. Cortical thickness analyses showed damage to the left lateral temporal cortex in the entire PPA cohort suggesting that lexical components are anatomically segregated within this cortical region. Our results also refine previous proposals about lexical deficits in PPA by demonstrating differential damage to all three components of the lexicon in semantic and logopenic PPA, which might have an impact on PPA diagnosis and language rehabilitation strategies. HighlightsAn organization of the lexicon is proposed by studying patients with primary aphasia.Priming tasks showed a segregation in semantic/syntactic/word‐form components.Cortical thickness analyses indicated their implementation in lateral temporal cortices.Patients with primary aphasia have differential damage to all lexical components.

Statistical Learning of Spatiotemporal Patterns from Longitudinal Manifold-Valued Networks

We introduce a mixed-effects model to learn spatiotempo-ral patterns on a network by considering longitudinal measures distributed on a fixed graph. The data come from repeated observations of subjects at different time points which take the form of measurement maps distributed on a graph such as an image or a mesh. The model learns a typical group-average trajectory characterizing the propagation of measurement changes across the graph nodes. The subject-specific trajectories are defined via spatial and temporal transformations of the group-average scenario, thus estimating the variability of spatiotemporal patterns within the group. To estimate population and individual model parameters, we adapted a stochastic version of the Expectation-Maximization algorithm, the MCMC-SAEM. The model is used to describe the propagation of cortical atrophy during the course of Alzheimers Disease. Model parameters show the variability of this average pattern of atrophy in terms of trajectories across brain regions, age at disease onset and pace of propagation. We show that the personaliza-tion of this model yields accurate prediction of maps of cortical thickness in patients.

Cerebral microbleeds and CSF Alzheimer biomarkers in primary progressive aphasias

Objective To reveal the prevalence and localization of cerebral microbleeds (CMBs) in the 3 main variants of primary progressive aphasia (PPA) (logopenic, semantic, and nonfluent/agrammatic), to identify the relationship with underlying Alzheimer pathology, and to explore whether CMBs contribute to language breakdown. Methods We used a cross-sectional design in a multicenter cohort of 82 patients with PPA and 19 similarly aged healthy controls. MRI allowed for rating CMBs (2-dimensional gradient recalled echo T2*, susceptibility weighted imaging sequences) and white matter hyperintensities. CSF Alzheimer disease biomarker analyses available in 63 of the 82 patients provided the stratification of PPA into subgroups with patients who had or did not have probable underlying Alzheimer pathology. Results The prevalence of CMBs was higher in patients with PPA (28%) than in controls (16%). They were more prevalent in logopenic PPA (50%) than in semantic PPA (18%) and nonfluent/agrammatic PPA (17%). The localization of CMBs was mainly lobar (81%) with no difference between the PPA variants. CMBs were more frequent in PPA patients with positive than with negative CSF Alzheimer disease biomarkers (67% vs 20%). Patients with and without lobar CMBs had similar volumes of white matter hyperintensities. Language and general cognitive impairment in PPA was unrelated to CMB rates. Conclusions CMB prevalence in PPA is higher than in healthy controls. CMBs were most prevalent in the logopenic variant, were related to underlying Alzheimer pathology, and did not affect the language/cognitive impairment. Our findings also suggest that CMB detection with MRI contributes to PPA variant diagnosis, especially of logopenic PPA, and provides an estimator of the underlying neuropathology.