Functional connectivity Diseases of connectivity Gwenalle Douaud FMRIB
Functional connectivity: Diseases of connectivity Gwenaëlle Douaud FMRIB, University of Oxford ISMRM Educational course – 10 th of May 2014 1/33
Diseases of connectivity or disconnection? • Lesion/degeneration/synaptic malfunction structural connectivity functional connectivity (e. g. , Cabral et al. , 2012): Abnormal functional connectivity in depression chronic pain Parkinson’s Alzheimer’s schizophrenia • Functional connectivity impairment disconnection syndrome, where “damage to the connection results in deficit that is dinstinct both from damage to the target and source regions” (Kleinschmidt & Vuilleumier, 2013) Gerstmann syndrome: acalculia +finger agnosia +left-right disorientation +agraphia Rusconi et al. , 2009 ISMRM Educational course – 10 th of May 2014 2/33
Resting-state f. MRI: advantages • Increased signal-to-noise ratio (Fox & Greicius, Review 2010): - at best, task-related modulation explains 20% of BOLD variance - spontaneous ongoing activity explains 50 -80% of BOLD variance ISMRM Educational course – 10 th of May 2014 3/33
Resting-state f. MRI: advantages • Covers the entire repertoire of functional networks used by the brain in “action” (Smith et al. , 2009) RSN: 36 healthy subjects f. MRI: ~7, 300 maps, ~30, 000 subjects ISMRM Educational course – 10 th of May 2014 4/33
Resting-state f. MRI: advantages • Allows for a broader sampling of patient populations asleep, sedated, too impaired for task-based f. MRI scanning, etc. Greicius et al. , 2008 • Is not confounded by task performance, effort, practice effects, etc. ISMRM Educational course – 10 th of May 2014 5/33
Resting-state f. MRI: inconvenients • “Rest” is a task state in itself, with potential performance differences, rather than differences in the underlying, stable brain organisation (Buckner et al. , 2008, 2013) Might still reveal some meaningful differences, just need careful interpretation • More susceptible to movement confounds: add motion parameters as covariate use ICA+FIX (automatic denoising using FSL tools: Salimi-Khorshidi et al. , 2014, Griffanti et al. , 2014) ISMRM Educational course – 10 th of May 2014 6/33
Resting-state f. MRI: inconvenients • Interpretation: - no causality information (yet) effective functional connectivity - no easy interpretation what (a change in) + and – correlations mean Smith et al. , 2013 ISMRM Educational course – 10 th of May 2014 7/33
Resting-state f. MRI in disease: reviews • Mild cognitive impairment/Alzheimer’s disease: - Dennis & Thompson, 2014 - Sheline & Raichle, 2013 • Movement disorders (esp. Parkinson’s disease): - Poston & Eidelberg, 2012 • Psychiatric disorders (e. g. , schizophrenia, ADHD, autism): - Greicius, 2008 - Posner et al. , 2014 ISMRM Educational course – 10 th of May 2014 8/33
Resting-state f. MRI analysis: seed-based approach in Parkinson’s disease • Seed-based approach - a priori knowledge/hypothesis Parkinson’s disease: Helmich et al. , 2010 ISMRM Educational course – 10 th of May 2014 9/33
Resting-state f. MRI analysis: seed-based approach in Parkinson’s disease • Seed-based approach - a priori knowledge/hypothesis Parkinson’s disease: Helmich et al. , 2010 ISMRM Educational course – 10 th of May 2014 10/33
Resting-state f. MRI analysis: seed-based approach in Parkinson’s disease • Seed-based approach - a priori knowledge/hypothesis Parkinson’s disease: Helmich et al. , 2010 Functional compensation with anterior putamen “taking over” connections to IPC: increased connectivity between IPC and anterior putamen in Parkinson’s was larger for the least-affected side • Very careful study: - negative control with DMN - corrected for motion (higher in patients) - checked for the effect of tremor: no tremor versus tremor spatial map, regressing out muscle activity (electromyography) - checked effect of medication - checked for grey matter volume differences of seeds and whole-brain VBM ISMRM Educational course – 10 th of May 2014 11/33
Resting-state f. MRI analysis: ICA-based approach in Alzheimer’s disease • ICA-based approach – more exploratory (though can also be hypothesis-driven) Alzheimer’s disease: Zamboni et al. , 2013 Dual regression for group comparisons ISMRM Educational course – 10 th of May 2014 12/33
Resting-state f. MRI analysis: ICA-based approach in Alzheimer’s disease • ICA-based approach – more exploratory (though can also be hypothesis-driven) Alzheimer’s disease: Zamboni et al. , 2013 ISMRM Educational course – 10 th of May 2014 13/33
Resting-state f. MRI analysis: ICA-based approach in Alzheimer’s disease • ICA-based approach – more exploratory (though can also be hypothesis-driven) Alzheimer’s disease: Zamboni et al. , 2013 Resting-state f. MRI less confounds, task f. MRI more interpretable: “Increased frontal activity during a memory task overlaps with increased frontal connectivity during rest in AD patients, suggesting that residual cognitive ability can be assessed using resting f. MRI. ” • Very careful study: - same number of healthy and AD participants for ICA - negative control with auditory RSN - corrected for GM volume - checked for the effect of physiological fluctuations (respiratory + cardiac activity) ISMRM Educational course – 10 th of May 2014 14/33
Resting-state f. MRI analysis: Graph-based approach in schizophrenia • Graph theory – exploratory (though mostly no basal ganglia or cerebellum) Schizophrenia: van den Heuvel et al. , 2013 ISMRM Educational course – 10 th of May 2014 15/33
Resting-state f. MRI analysis: Graph-based approach in schizophrenia • Graph theory – exploratory (though mostly no basal ganglia or cerebellum) Schizophrenia: van den Heuvel et al. , 2013 “Reduced level of rich club interconnectivity in patients with schizophrenia (…), potentially resulting in decreased global communication capacity and altered functional brain dynamics” • Careful study: - includes basal ganglia - used Freesurfer parcellation for ROIs (as opposed to AAL) - replication dataset effects not specific to Rich Club - but: “This study did not reveal a clear association between clinical metrics of patients and rich club organization” ISMRM Educational course – 10 th of May 2014 16/33
Resting-state f. MRI analysis: Multi-modal approach in motor neuron disease • Combining information – diffusion tensor and tractography Amyotrophic lateral sclerosis: Douaud, Filippini et al. , 2011 Increase FC in ALS ISMRM Educational course – 10 th of May 2014 17/33
Resting-state f. MRI analysis: Multi-modal approach in motor neuron disease • Combining information – diffusion tensor and tractography Amyotrophic lateral sclerosis: Douaud, Filippini et al. , 2011 • Careful registration (BBR + VBM) Disease duration ISMRM Educational course – 10 th of May 2014 18/33
Resting-state f. MRI analysis: Multi-modal approach in motor neuron disease • Combining information – diffusion tensor and tractography Amyotrophic lateral sclerosis: Douaud, Filippini et al. , 2011 Higher functional connectivity not necessarily better • Reconciling lower structural connectivity (SC) with higher functional connectivity? corpus callosum GABAergic interneurons Innocenti, 2009 ISMRM Educational course – 10 th of May 2014 19/33
Resting-state f. MRI analysis: Multi-modal approach in motor neuron disease • Combining information – diffusion tensor and tractography Amyotrophic lateral sclerosis: Douaud, Filippini et al. , 2011 Low SC + high FC in ALS = loss of GABA interneurons + FC ISMRM Educational course – 10 th of May 2014 - GABA 20/33
Resting-state f. MRI analysis: Multi-modal approach in neurodegenerative diseases • Combining information – grey matter volume/structural covariance Array of neurodegenerative disorders: Seeley et al. , 2009 ISMRM Educational course – 10 th of May 2014 21/33
Resting-state f. MRI analysis: Multi-modal approach in neurodegenerative diseases • Combining information – grey matter volume/structural covariance Array of neurodegenerative disorders: Seeley et al. , 2009 Dissociable networks for each disease ISMRM Educational course – 10 th of May 2014 22/33
Variability of results in fc. MRI Fox & Greicius, 2010 ISMRM Educational course – 10 th of May 2014 23/33
Variability of results in fc. MRI: some guidelines Parkinson’s: Seeds in the striatum DMN as negative control Alzheimer’s: RSN (ICA) involving frontal areas auditory RSN as negative control Fox & Greicius, 2010 ISMRM Educational course – 10 th of May 2014 24/33
Variability of results in fc. MRI: some guidelines + careful registration Fox & Greicius, 2010 ISMRM Educational course – 10 th of May 2014 25/33
Variability of results in fc. MRI: some guidelines + careful registration Fox & Greicius, 2010 ISMRM Educational course – 10 th of May 2014 26/33
Variability of results in fc. MRI: movement “Scrub” the data, add motion parameters, or use ICA+FIX Power et al. , 2012 ISMRM Educational course – 10 th of May 2014 27/33
Variability of results in fc. MRI: movement “Scrub” the data, add motion parameters, or use ICA+FIX Salimi-Khorshidi et al. , 2014 Griffanti et al. , 2014 ISMRM Educational course – 10 th of May 2014 28/33
Variability of results in fc. MRI: some guidelines Global signal regression, # of ICs etc. Fox & Greicius, 2010 ISMRM Educational course – 10 th of May 2014 29/33
Variability of results in fc. MRI: some guidelines Fox & Greicius, 2010 ISMRM Educational course – 10 th of May 2014 30/33
Variability of results in fc. MRI: stability of networks • Inter-subject variability is higher in higher-order regions (Mueller et al. , 2013) ISMRM Educational course – 10 th of May 2014 31/33
Interpretation of functional connectivity results • Some RSN are more stable than others • Higher not necessarily better • Always check for each contrast what happens in each cluster Absolute values of correlations matter ISMRM Educational course – 10 th of May 2014 32/33
Interpretation of functional connectivity results • Some RSN are more stable than others • Higher not necessarily better • Always check for each contrast what happens in each cluster It’s the absolute values of correlations that matter • Bear in mind that change in correlations can be observed even in the absence of a change in coupling (Friston, 2011) Changes in correlation between A and B could be caused by a change in correlation elsewhere Changes in correlation could be caused by a change in SNR (e. g. , heart rate variability differs between two populations) Changes in correlation could be caused by a change in the amplitude of the fluctuations • Bear in mind that “resting” is to some extent also a task ISMRM Educational course – 10 th of May 2014 33/33
Special thanks to: FMRIB, University of Oxford - Steve Smith - Eugene Duff - Christian Beckmann - Reza Salimi-Khorshidi - Martin Turner - Giovanna Zamboni - Nicola Filippini - Marina Charquero Ballester THANK YOU FOR YOUR ATTENTION ISMRM Educational course – 10 th of May 2014 34/33
- Slides: 34