Investigating microstructural variation in the human hippocampus using non-negative matrix factorization

Steele, Christopher; Chen, Anthony G. X.; Patel, Sejal; Devenyi, Gabriel A.; Germann, Jürgen; Tardif, Christine; Chakravarty, M. Mallar

doi:10.1016/j.neuroimage.2019.116348

Cited by 53 publications

(113 citation statements)

References 130 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Diffusion weighted imaging derived maps of FA, MD, and RD were co-registered to the T1-weighted image to allow for mid-surface sampling of these metrics. For each of the five microstructural metrics a 77122 x 398 (number of vertices x number of subjects) matrix was constructed, each of which was then concatenated to build a 77122 x 1990 (number of vertices x (number of subjects*5) multimodal input matrix which was input to NMF (24, 26). NMF is a decomposition technique which identifies spatial components and subject weightings, together identifying regions of the brain where microstructural variation is observed (spatial components) as well as each individual’s microstructural profile in a given component (subject weightings).…”

Section: Resultsmentioning

confidence: 99%

“…By modelling shared covariance across MRI metrics, as opposed to separately analysing each piece of information, this allows for a more comprehensive assessment of differences across subjects (23). To this end, we use non-negative matrix factorization (NMF), a matrix decomposition technique previously used in our group to probe microstructural properties of the hippocampus (24). Applied to MRI data, NMF highlights regions of the brain in which shared patterns of microstructural variation occur, as well as subject-specific measurements describing individual microstructural features within the highlighted brain regions.…”

Section: Introductionmentioning

confidence: 99%

“…Applied to MRI data, NMF highlights regions of the brain in which shared patterns of microstructural variation occur, as well as subject-specific measurements describing individual microstructural features within the highlighted brain regions. (24–26).…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Individual variation in brain structural-cognition relationships in aging

Mackay

Jansen²,

Devenyi

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

While all individuals are susceptible to age-related cognitive decline, significant inter- and intra-individual variability exists. However, the sources of this variation remain poorly understood. Here, we examined the association between 30-year trajectories of cognitive decline and multimodal indices of brain microstructure and morphology in older age. We used the Whitehall II Study, an extensively characterised cohort using 3T brain magnetic resonance images acquired at older age (mean age = 69.52 4.9) and 5 repeated cognitive performance assessments between mid-life (mean age = 53.2 4.9 years) and late-life (mean age = 67.7 4.9). Using non-negative matrix factorization, we identified 10 brain microstructural components that integrate measures of cortical thickness, surface area, fractional anisotropy, and mean and radial diffusivities. We observed two modes of variance that describe the association between cognition and brain microstructure. The first describes variations in 5 microstructural components associated with low mid-life performance across multiple cognitive domains, decline in reasoning abilities, but a relative maintenance of lexical and semantic fluency from mid-to-late life. The second describes variations in 5 microstructural components that are associated with low mid-life performance in lexical fluency, semantic fluency and short-term memory performance, but a retention of abilities in multiple domains from mid-to-late life. The extent to which a subject loads onto a latent variables predicts their future cognitive performance 3.2 years later (mean age = 70.87 4.9). This data-driven approach highlights a complex pattern of brain-behavior relationships, wherein the same individuals express both decline and maintenance in function across cognitive domains and in brain structural features.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Individual variation in brain structural-cognition relationships in aging

Mackay

Jansen²,

Devenyi

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Previously, medial/lateral (ML) gradient was found to correlate with an estimation of myelin content in the hippocampus, with the medial hippocampus having a higher myelin content (Vos de Wael et al 2018). Further, a recent paper using a non-negative matrix factorization technique to cluster the hippocampus demonstrated that the medial hippocampus cluster was the most myelinated (Patel et al2020). Similarly, intracortical myelin was found to be the highest in the subiculum, which is the most medial hippocampal subfield (DeKraker et al 2018).…”

Section: Discussionmentioning

confidence: 99%

“…Our "cognitive/demographic" data contained age, sex, years of education, APOE4 status, and RBANS scores for each subject (matrix size 85x5). Following a statistical protocol described in previous works (McIntosh and Lobaugh 2004;Krishnan et al 2011;McIntosh and Mišić 2013;Zeighami et al 2017;Patel et al 2020), each LV was then tested statistically using permutation testing. First, row permutations (10,000 times) of the input "brain" matrix were subject to PLS in order to obtain a distribution of singular values with the hypothesis that a permuted "brain" matrix will eliminate the initial brain-cognitive relationships.…”

Section: 92partial Least Squares Analysismentioning

confidence: 99%

Hippocampus shape across the healthy lifespan and its relationship with cognition

Bussy

Plitman

Tullo

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

The study of the hippocampus across the healthy adult lifespan has rendered inconsistent findings. While volumetric measurements have often been a popular technique for analysis, more advanced morphometric techniques have demonstrated compelling results that highlight the importance and improved specificity of shape-based measures. Here, the MAGeT Brain algorithm was applied on 134 healthy individuals aged 18-81 years old to extract hippocampal subfield volumes and hippocampal shape measurements, notably : local surface area (SA) and displacement. We used linear, second or third order natural splines to examine the relationships between hippocampal measures and age. In addition, partial least squares analyses were performed to relate measurements with cognitive and demographic information. Volumetric results indicated a relative preservation of the right cornus ammonis 1 with age and a global volume reduction linked with older age, female sex, lower levels of education and cognitive performance. Vertex-wise analysis demonstrated an SA preservation in the anterior hippocampus with a peak during the sixth decade, while the posterior hippocampal SA gradually decreased across lifespan. Overall, SA decrease was linked to older age, female sex and, to a lesser extent lower levels of education and cognitive performance. Outward displacement in the lateral hippocampus and inward displacement in the medial hippocampus were enlarged with older age, lower levels of cognition and education, indicating an accentuation of the hippocampal C shape with age. Taken together, our findings suggest that vertex-wise analyses have higher spatial specifity and that sex, education and cognition are implicated in the differential impact of age on hippocampal subregions throughout its antero-posterior and medial-lateral axes.

show abstract

Data‐driven decomposition and staging of flortaucipir uptake in Alzheimer's disease

Earnest,

Bani,

et al. 2024

Alzheimer's & Dementia

View full text Add to dashboard Cite

INTRODUCTIONPrevious approaches pursuing in vivo staging of tau pathology in Alzheimer's disease (AD) have typically relied on neuropathologically defined criteria. In using predefined systems, these studies may miss spatial deposition patterns which are informative of disease progression.METHODSWe selected discovery (n = 418) and replication (n = 132) cohorts with flortaucipir imaging. Non‐negative matrix factorization (NMF) was applied to learn tau covariance patterns and develop a tau staging system. Flortaucipir components were also validated by comparison with amyloid burden, gray matter loss, and the expression of AD‐related genes.RESULTSWe found eight flortaucipir covariance patterns which were reproducible and overlapped with relevant gene expression maps. Tau stages were associated with AD severity as indexed by dementia status and neuropsychological performance. Comparisons of flortaucipir uptake with amyloid and atrophy also supported our model of tau progression.DISCUSSIONData‐driven decomposition of flortaucipir uptake provides a novel framework for tau staging which complements existing systems.Highlights NMF reveals patterns of tau deposition in AD. Data‐driven staging of flortaucipir tracks AD severity. Learned flortaucipir patterns overlap with AD‐related gene expression.

show abstract

Investigating microstructural variation in the human hippocampus using non-negative matrix factorization

Cited by 53 publications

References 130 publications

Individual variation in brain structural-cognition relationships in aging

Individual variation in brain structural-cognition relationships in aging

Hippocampus shape across the healthy lifespan and its relationship with cognition

Data‐driven decomposition and staging of flortaucipir uptake in Alzheimer's disease

Contact Info

Product

Resources

About