On the stability of canonical correlation analysis and partial least squares with application to brain-behavior associations

Communications Biology, Год журнала: 2024, Номер 7(1)

Опубликована: Фев. 21, 2024

Abstract Associations between datasets can be discovered through multivariate methods like Canonical Correlation Analysis (CCA) or Partial Least Squares (PLS). A requisite property for interpretability and generalizability of CCA/PLS associations is stability their feature patterns. However, in high-dimensional questionable, as found empirical characterizations. To study these issues systematically, we developed a generative modeling framework to simulate synthetic datasets. We that when sample size relatively small, but comparable typical studies, are highly unstable inaccurate; both magnitude importantly the pattern underlying association. confirmed trends across two neuroimaging modalities independent with n ≈ 1000 = 20,000, only latter comprised sufficient observations stable mappings imaging-derived behavioral features. further power calculator provide sizes required reliability analyses. Collectively, characterize how limit detrimental effects overfitting on stability, recommendations future studies.

Язык: Английский

---

Functional magnetic resonance imaging in schizophrenia: current evidence, methodological advances, limitations and future directions

World Psychiatry, Год журнала: 2024, Номер 23(1), С. 26 - 51

Опубликована: Янв. 12, 2024

Functional neuroimaging emerged with great promise and has provided fundamental insights into the neurobiology of schizophrenia. However, it faced challenges criticisms, most notably a lack clinical translation. This paper provides comprehensive review critical summary literature on functional neuroimaging, in particular magnetic resonance imaging (fMRI), We begin by reviewing research fMRI biomarkers schizophrenia high risk phase through historical lens, moving from case-control regional brain activation to global connectivity advanced analytical approaches, more recent machine learning algorithms identify predictive features. Findings studies negative symptoms as well neurocognitive social cognitive deficits are then reviewed. neural markers these may represent promising treatment targets Next, we summarize related antipsychotic medication, psychotherapy psychosocial interventions, neurostimulation, including response resistance, therapeutic mechanisms, targeting. also utility data-driven approaches dissect heterogeneity schizophrenia, beyond comparisons, methodological considerations advances, consortia precision fMRI. Lastly, limitations future directions field discussed. Our suggests that, order for be clinically useful care patients should address potentially actionable decisions that routine treatment, such which prescribed or whether given patient is likely have persistent impairment. The potential influenced must weighed against cost accessibility factors. Future evaluations prognostic consider health economics analysis.

Язык: Английский

---

Two common and distinct forms of variation in human functional brain networks

Nature Neuroscience, Год журнала: 2024, Номер 27(6), С. 1187 - 1198

Опубликована: Апрель 30, 2024

Язык: Английский

---

On stability of Canonical Correlation Analysis and Partial Least Squares with application to brain-behavior associations

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2020, Номер unknown

Опубликована: Авг. 25, 2020

Abstract Associations between datasets can be discovered through multivariate methods like Canonical Correlation Analysis (CCA) or Partial Least Squares (PLS). A requisite property for interpretability and generalizability of CCA/PLS solutions is stability feature patterns driving an association. However, in high-dimensional questionable, as found empirical characterizations. To study these issues a systematic manner, we developed generative modeling framework to simulate synthetic datasets, parameterized by dimensionality, variance structure, association strength. We that when sample size relatively small, but comparable typical studies, associations are highly unstable inaccurate; both their magnitude importantly the latent pattern underlying confirmed trends across two neuroimaging modalities, functional diffusion MRI, independent Human Connectome Project (n ≈ 1000) UK Biobank 20000) only latter comprised sufficient samples stable mappings imaging-derived behavioral features. further power calculator provide sizes required reliability analyses future studies.

Язык: Английский

---

Functional neuroimaging in psychiatry and the case for failing better

Neuron, Год журнала: 2022, Номер 110(16), С. 2524 - 2544

Опубликована: Авг. 1, 2022

Язык: Английский

---

The Burden of Reliability: How Measurement Noise Limits Brain-Behaviour Predictions

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2023, Номер unknown

Опубликована: Фев. 10, 2023

Abstract Major efforts in human neuroimaging strive to understand individual differences and find biomarkers for clinical applications by predicting behavioural phenotypes from brain imaging data. An essential prerequisite identifying generalizable replicable brain-behaviour prediction models is sufficient measurement reliability. However, the selection of targets predominantly guided scientific interest or data availability rather than reliability considerations. Here we demonstrate impact low phenotypic on out-of-sample performance. Using simulated empirical Human Connectome Projects, found that levels common across many can markedly limit ability link behaviour. Next, using 5000 subjects UK Biobank, show only highly reliable fully benefit increasing sample sizes hundreds thousands participants. Overall, our findings highlight importance brain–behaviour associations differences.

Язык: Английский

---

Replicable brain–phenotype associations require large-scale neuroimaging data

Nature Human Behaviour, Год журнала: 2023, Номер 7(8), С. 1344 - 1356

Опубликована: Июнь 26, 2023

Язык: Английский

---

Microstructural and functional plasticity following repeated brain stimulation during cognitive training in older adults

Nature Communications, Год журнала: 2023, Номер 14(1)

Опубликована: Июнь 2, 2023

Abstract The combination of repeated behavioral training with transcranial direct current stimulation (tDCS) holds promise to exert beneficial effects on brain function beyond the trained task. However, little is known about underlying mechanisms. We performed a monocenter, single-blind randomized, placebo-controlled trial comparing cognitive concurrent anodal tDCS (target intervention) sham (control intervention), registered at ClinicalTrial.gov (Identifier NCT03838211). primary outcome (performance in task) and secondary outcomes transfer tasks) were reported elsewhere. Here, mechanisms addressed by pre-specified analyses multimodal magnetic resonance imaging before after three-week executive prefrontal 48 older adults. Results demonstrate that combined active modulated white matter microstructure which predicted individual task performance gain. Training-plus-tDCS also resulted microstructural grey alterations site, increased functional connectivity. provide insight into neuromodulatory interventions, suggesting tDCS-induced changes fiber organization myelin formation, glia-related synaptic processes target region, synchronization within targeted networks. These findings advance mechanistic understanding neural effects, thereby contributing more network modulation future experimental translation applications.

Язык: Английский

---

Reply to: Multivariate BWAS can be replicable with moderate sample sizes

Nature, Год журнала: 2023, Номер 615(7951), С. E8 - E12

Опубликована: Март 8, 2023

Язык: Английский

---

The functional connectome in obsessive-compulsive disorder: resting-state mega-analysis and machine learning classification for the ENIGMA-OCD consortium

Molecular Psychiatry, Год журнала: 2023, Номер 28(10), С. 4307 - 4319

Опубликована: Май 2, 2023

Abstract Current knowledge about functional connectivity in obsessive-compulsive disorder (OCD) is based on small-scale studies, limiting the generalizability of results. Moreover, majority studies have focused only predefined regions or networks rather than throughout entire brain. Here, we investigated differences resting-state between OCD patients and healthy controls (HC) using mega-analysis data from 1024 1028 HC 28 independent samples ENIGMA-OCD consortium. We assessed group whole-brain at both regional network level, whether could serve as biomarker to identify patient status individual level machine learning analysis. The mega-analyses revealed widespread abnormalities OCD, with global hypo-connectivity (Cohen’s d : -0.27 -0.13) few hyper-connections, mainly thalamus 0.19 0.22). Most hypo-connections were located within sensorimotor no fronto-striatal found. Overall, classification performances poor, area-under-the-receiver-operating-characteristic curve (AUC) scores ranging 0.567 0.673, better for medicated (AUC = 0.702) unmedicated 0.608) versus controls. These findings provide partial support existing pathophysiological models highlight important role OCD. However, does not so far an accurate identifying level.

Язык: Английский

---