Presentation is loading. Please wait.

Presentation is loading. Please wait.

Susana Muñoz Maniega Research Fellow, Disconnected Mind Project SFC Brain Imaging Research Centre Division of Clinical Neurosciences University of Edinburgh.

Similar presentations


Presentation on theme: "Susana Muñoz Maniega Research Fellow, Disconnected Mind Project SFC Brain Imaging Research Centre Division of Clinical Neurosciences University of Edinburgh."— Presentation transcript:

1 Susana Muñoz Maniega Research Fellow, Disconnected Mind Project SFC Brain Imaging Research Centre Division of Clinical Neurosciences University of Edinburgh From water random motion to brain's white matter fibres and the study of cognition

2 Overview Water diffusivity in the brain White matter integrity biomarkers Whole brain analysis – voxel-based Tractography methods LBC1936 – white matter and cognition Role of computational resources

3 Diffusion MRI: Background Diffusion is the random translational motion (Brownian motion) due to thermal energy In tissues, diffusivity is affected by the local cellular environment If the cell membranes have directional coherence, then diffusion will depend on direction – anisotropic diffusion Robert Brown 1773 - 1858 Albert Einstein 1879 - 1955

4 Diffusion MRI: Background Diffusion is the random translational motion (Brownian motion) due to thermal energy In tissues, diffusivity is affected by the local cellular environment If the cell membranes have directional coherence, then diffusion will depend on direction – anisotropic diffusion

5 Diffusion MRI: Background Diffusion is the random translational motion (Brownian motion) due to thermal energy In tissues, diffusivity is affected by the local cellular environment If the cell membranes have directional coherence, then diffusion will depend on direction – anisotropic diffusion Diffusion perpendicular to long axis Diffusion parallel to long axis

6 Imaging biomarkers Mean diffusivity (MD = mean{λi=1,3}) magnitude of diffusion Fractional anisotropy (FA = var{λi=1,3}/magn{D}) directional coherence: 0 indicates isotropic diffusion (CSF) 1 indicates highly anisotropic diffusion (white matter) MD FA Healthy, structurally intact white matter has low MD and high FA Structurally compromised white matter has high MD and low FA

7 A Voxel-Based Analysis Approach We can look for correlations of FA with other parameters in a hypothesis-free manner looking at the whole brain white matter Tract-based spatial statistics (TBSS) is a voxel-based analysis approach customised for the study of diffusion parameters in white matter Aligned Averaged Thinned FA projected into skeleton Stats Smith et al. NeuroImage 2006 31:1487-1505

8 TBSS In VBA the accurate registration is crucial – usually all brains are registered to a brain template For a cohort of older subjects we cannot use templates (created from younger brains) so we chose a registration target from the database itself as the most typical This minimises the registration errors, but at the cost of time TBSS preprocessing requires N ×N registrations each taking ~ 5 min http://www.fmrib.ox.ac.uk/fsl/tbss/

9 White matter integrity and age 90 subjects 65 to 88 years old 90 ×90 registrations ~ 28 days 1-2 days in parallel Widespread negative correlations between FA and age p < 0.05

10 FA Tractography

11 Reconstruct white matter tracts in 3D by piecing together voxel-based estimates of the underlying continuous fibre orientation field Mori et al. NMR Biomed 2002 15:468-480

12 Behrens et al. NeuroImage 2007 34:144-155 We use probabilistic diffusion tractography (Bedpostx/Probtrackx) with a model for fitting 2 fibre orientations in each voxel To perform tractography in a group study we need to automatize the process but still segmenting the tracts reliably in all subjects Tractography

13 Neighbourhood Tractography NT selects a seed point from the set of candidates using a reference tract as a guide to the expected topology of the segmented tract –NT models the variability in shape and length of the tract and finds the tract that best matches the model from a set of candidates –An EM algorithm is used to fit the model Clayden et al. NeuroImage 2009 45:377-385

14 same tract is segmented in each brain http://code.google.com/p/tractor/

15 The Lothian Birth Cohort 1936 (LBC1936) comprises 1091 surviving participants of the Scottish Mental Survey 1947 (SMS1947) who now live in the Lothian area of Scotland They were recruited at age about 70 into a follow-up study The childhood cognitive ability data provide a baseline from which to calculate life-long cognitive changes Deary et al. BMC Geriatrics 2007, 7:28 2007 1947 LBC1936

16 Using contemporary brain MRI (at age 72-73), including diffusion tensor imaging (DTI), we examined how white matter integrity relates to changes in cognition in the LBC1936 We used fractional anisotropy (FA) and mean diffusivity (MD) as markers of white matter integrity in specific tracts White matter integrity was related to IQ (11 and 70) and general factors of cognition, speed and memory MRI role in the DM Project

17 Tracts of interest

18 N=318 Preliminary results show association between uncinate fasciculus integrity and intelligence at ages 11 and 70 This supports the hypothesis that uncinate fasciculus is part of the neural basis for intelligence Tracts and cognition

19 Storage: –Raw data ~ 375 MB –Pre-processed tractography dataset ~ 500 MB –Pre-processed structural dataset ~ 250 MB Total × 1000 ~ 1TB Processing in a single computer: –Diffusion data pre-processing ~ 20 min × 1000 = 13.8 days –Data modelling for tractography (BedpostX with 2 fibre model) > 24 h per dataset × 1000 = 2.7 years –NT tract shape modelling ~ 2.5 h per subject per tract × 1000 = 3.5 months × 14 tracts = 4.1 years Total: ~ 7 years … computational issues in an imaging study of 1000 subjects

20 The TractoR (Tractography with R) project includes R packages for processing, analysing and visualising magnetic resonance images (available in CRAN) R based scripting infrastructure and shell script frontend for running common analyses Facilitates running R code on parallelised systems Configurable with design files containing options, e.g. imaging datasets to analyse in parallel Processing of LBC1936 data in Eddie –200 datasets processed in ~ 1-2 weeks http://code.google.com/p/tractor/

21 Disconnected mind, and particularly LBC1936, is an unique study that would give insights into normal cognitive ageing The size of the cohort is both a strength and a weakness Analysis of all the data is only possible using parallelisable processes and the Edinburgh Compute and Data Facility Conclusions

22 Acknowledgements www.disconnectedmind.ed.ac.uk This work has made use of the resources provided by the Edinburgh Compute and Data Facility (ECDF). (http://www.ecdf.ed.ac.uk). The ECDF is partially supported by the eDIKT initiative. (http://www.edikt.org) Ian J. DearyKaren Horsburgh James McCulloch Richard MorrisJohn StarrJoanna Wardlaw Mark BastinEmma WoodIan Marshall Principal Investigators Research team: Caroline BrettRobin ColtmanJanie Corley Stephanie DaumasGail DaviesPaula Davies Ruth DeightonTommy DingwallJill Fowler Catherine GliddonAlan GowSarah Harris Ross HendersonPhilip HollandLorna Houlihan Karim KhalloutSeverine LaunayMichelle Luciano Kevin McGheeCatherine MurrayAlison Pattie Lars PenkePaul RedmondMichell Reimer Natalie RoyleFiona ScottJessica Smith Aisling SpainYanina Tsenkina Maria Valdés Hernández Susana Muñoz Maniega


Download ppt "Susana Muñoz Maniega Research Fellow, Disconnected Mind Project SFC Brain Imaging Research Centre Division of Clinical Neurosciences University of Edinburgh."

Similar presentations


Ads by Google