Machine learning in resting-state fMRI analysis

Khosla, Meenakshi; Jamison, Keith; Ngo, Gia H.; Kuceyeski, Amy; Sabuncu, Mert R.

doi:10.1016/j.mri.2019.05.031

Cited by 151 publications

(88 citation statements)

References 167 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For each segment, fit a linear function using least squares to determine slope, 6 and intercept, 6 . Subtract best fitting linear trend and compute fluctuation from mean of resulting signal, 6 To avoid biases introduced by volumetric sex differences, 390 subjects (190 nonoverlapping male-female pairs) were selected such that each pair had a matched total grey matter volume (cortical and subcortical) with a percent difference less than or equal to 1%. The final volume-matched sample did not differ in total grey matter volume (p>0.05) between males and females.…”

Section: Hurst Exponentmentioning

confidence: 99%

See 1 more Smart Citation

Sex classification using long-range temporal dependence of resting-state functional MRI time series

Dhamala

Jamison

Sabuncu

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

A thorough understanding of sex differences, if any, that exist in the brains of healthy individuals is crucial for the study of neurological illnesses that exhibit differences in clinical and behavioural phenotypes between males and females. In this work, we evaluate sex differences in regional temporal dependence of resting-state brain activity using 195 male-female pairs (aged 22-37) from the Human Connectome Project. Malefemale pairs are strictly matched for total grey matter volume. We find that males have more persistent long-range temporal dependence than females in regions within temporal, parietal, and occipital cortices. Machine learning algorithms trained on regional temporal dependence measures achieve sex classification accuracies of up to 81%.Regions with the strongest feature importance in the sex classification task included cerebellum, amygdala, frontal cortex, and occipital cortex. Additionally, we find that even after males and females are strictly matched on total grey matter volume, significant regional volumetric sex differences persist in many cortical and subcortical regions. Our results indicate males have larger cerebella, hippocampi, parahippocampi, thalami, caudates, and amygdalae while females have larger cingulates, precunei, frontal cortices, and parietal cortices. Sex classification based on regional volume achieves accuracies of up to 85%; cerebellum, cingulate cortex, and temporal cortex are the most important features. These findings highlight the important role of strict volume matching when studying brain-based sex differences. Differential patterns in regional temporal dependence between males and females identifies a potential neurobiological substrate underlying sex differences in functional brain activation patterns and the behaviours with which they correlate.

show abstract

Section: Hurst Exponentmentioning

confidence: 99%

“…In recent years, machine learning techniques have increasingly been used in the analysis of resting-state fMRI data [6]. Supervised methods have been successfully applied to make subject-level predictions in both healthy and clinical populations [6].…”

Section: Introductionmentioning

confidence: 99%

Sex classification using long-range temporal dependence of resting-state functional MRI time series

Dhamala

Jamison

Sabuncu

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…More movies would not only increase generalisability, they would increase the number of stimulus features and events in a variety of (jittered) contexts that might be annotated. These could then be used to label finer grained patterns of activity, e.g., making machine learning/decoding approaches more feasible [55][56][57] .…”

Section: Natural-fmrimentioning

confidence: 99%

A ‘Naturalistic Neuroimaging Database’ for understanding the brain using ecological stimuli

Aliko

Huang

Gheorghiu

et al. 2020

Preprint

View full text Add to dashboard Cite

Neuroimaging has advanced our understanding of human psychology using reductionist stimuli that often do not resemble information the brain naturally encounters. It has improved our understanding of the network organization of the brain mostly through analyses of 'resting-state' data for which the functions of networks cannot be verifiably labelled. We make a ' Naturalistic Neuroimaging Database ' (NNDb v1.0) publically available to allow for a more complete understanding of the brain under more ecological conditions during which networks can be labelled. Eighty-six participants underwent behavioural testing and watched one of 10 full-length movies while functional magnetic resonance imaging was acquired. Resulting timeseries data are shown to be of high quality, with good signal-to-noise ratio, few outliers and low movement. Data-driven functional analyses provide further evidence of data quality. They also demonstrate accurate timeseries/movie alignment and how movie annotations might be used to label networks. The NNDb can be used to answer questions previously unaddressed with standard neuroimaging approaches, progressing our knowledge of how the brain works in the real world. judgement about these stimuli, with a corresponding button response (e.g., a '2AFC' indicating whether a sound is 'ba' or 'pa').The result of relying on unnatural stimuli and tasks is that our neurobiological understanding derived from task-fMRI may not be representative of how the brain processes information. This is perhaps why fMRI test-retest reliability is low 6,7 . Indeed, more ecologically valid stimuli like movies have higher reliability than resting-or task-fMRI. This is not only because these enhance activity, decrease head movement and improve participant compliance 8,9 . Rather, natural stimuli have higher test-retest reliability mostly because they are more representative of operations the brain normally performs and provide more constraints on processing 10-15 . Resting-fMRIThere has arguably been a significant increase in our understanding of the network organization of the human brain because of the public availability of large resting-fMRI datasets, analysed with dynamic and other functional connectivity methods 16,17 . These include the INDI '1000 Functional Connectomes Project' 18 , 'Human Connectome Project' (HCP) 19 and UK Biobank 20 . Collectively, these datasets have more than 6,500 participants sitting in a scanner 'resting'. Resulting resting-state networks are said to represent the 'intrinsic' network architecture of the brain, i.e., networks that are present even in the absence of exogenous tasks. These networks are often claimed to be modular and to constrain the task-based architecture of the brain 21 .As with task-fMRI, one might ask how representative resting-state networks are given that participants are anything but at rest. They are switching between fixating on a cross-hair, trying to stay awake, visualising, trying not to think and thinking through inner speech 21,22 . Some of these are not particular...

show abstract

“…Much of the research in this direction has aimed at identifying connectivity based biomarkers, restricting the analysis to so-called "static" functional connectivity measures that quantify the average degree of synchrony between brain regions. For e.g., machine learning based strategies have been used with static connectivity measures to parcellate the brain into functional networks, and extract individual-level predictions about cognitive state or clinical condition [2]. In recent years, there has been a surge in the study of the temporal dynamics of rs-fMRI data, offering a complementary perspective on the functional connectome and how it is altered in disease, development, and aging [14].…”

Section: Introductionmentioning

confidence: 99%

“…Related Work: Machine learning methods are increasingly used to compute individual-level predictions from rs-fMRI data, e.g. about disease [2]. The conventional approach of supervised learning relies on labeled training data and uses hand-crafted features such as the static correlation between pairs of regions.…”

Section: Introductionmentioning

confidence: 99%

Detecting Abnormalities in Resting-State Dynamics: An Unsupervised Learning Approach

Khosla

Jamison

Kuceyeski

et al. 2019

Machine Learning in Medical Imaging

Self Cite

View full text Add to dashboard Cite

Resting-state functional MRI (rs-fMRI) is a rich imaging modality that captures spontaneous brain activity patterns, revealing clues about the connectomic organization of the human brain. While many rs-fMRI studies have focused on static measures of functional connectivity, there has been a recent surge in examining the temporal patterns in these data. In this paper, we explore two strategies for capturing the normal variability in resting-state activity across a healthy population: (a) an autoencoder approach on the rs-fMRI sequence, and (b) a next frame prediction strategy. We show that both approaches can learn useful representations of rs-fMRI data and demonstrate their novel application for abnormality detection in the context of discriminating autism patients from healthy controls.

show abstract

Machine learning in resting-state fMRI analysis

Cited by 151 publications

References 167 publications

Sex classification using long-range temporal dependence of resting-state functional MRI time series

Sex classification using long-range temporal dependence of resting-state functional MRI time series

A ‘Naturalistic Neuroimaging Database’ for understanding the brain using ecological stimuli

Detecting Abnormalities in Resting-State Dynamics: An Unsupervised Learning Approach

Contact Info

Product

Resources

About