Presentation on theme: "Neuroimaging Processing : Overview, Limitations, pitfalls, etc. etc."— Presentation transcript:
1Neuroimaging Processing : Overview, Limitations, pitfalls, etc. etc.
2NeuroimagingNeuroimaging includes the use of various techniques to either directly or indirectly image the structure or function of the brain.Structural neuroimaging deals with the structure of the brain (e.g. shows contrast between different tissues: cerebrospinal fluid, grey matter, white matter).Functional neuroimaging is used to indirectly measure brain function (e.g. neural activity)Molecular neuroimaging measures biological processes in the brain at the molecular and cellular level.
5MRI Basics Water = H2O Each Hydrogen = one proton Protons Spin Generates detectable signal in externally applied magnetic field: that is, it causes protons to precess at a frequency proportional to the strength of the magnetic field – the ‘resonant’ frequencyWater Content ofGM 70%WM 85%Blood 93%Hydrogen AtomPROTON
7Magnetic Resonance Imaging (MRI) ExcitationRadio frequency (RF) pulse is applied at the precession frequency (Lamour Frequency)Sending an RF pulse at Lamour freq, particular amplitude and length of time – possible to flip the net magnetism 90° - perpendicular to Magnetic Field (B0)RelaxationT1-weighted is the time it takes for the protons to relax to B0Not all protons bound by their molecules in same way, dependant on tissue type
19VBM - Limitations Accuracy of the spatial normalisation Regular SPM uses 1000 parameters – just fits overall shape of the brain - mis-registrationsDeformation-based morphometry (e.g. DARTEL)– deformation field is analysedGrey matter matched with grey matter – doesn't’t indicate whether sulci/gyri are aligned
20FreeSurfer The cortex Volume, thickness or surface area? Volume = surface area * thickness
21Volume, thickness & surface area Related but don’t necessarily track each other ....Morphometry Differences between Young, Elderly and Mild Alzheimer’s in entorhinal cortex. *p<0.05Dickerson et al.2007
22Cortical Curvature Temporal Lobe Epilepsy (MR-negative) Cortical curvature abnormality in the ipsilateral temporal lobe - Not explained by volume or thicknessPossible surrogate marker for malformations of cortical developmentRonan et al. 2011
23FreeSurfer Cortical Reconstruction Cortical Analysis - cortical thickness, surface are, volume, cortical folding and curvatureCortical and sub-cortical segmentation
25Surface Model Mesh (“Finite Element”) Vertex = point of 6 triangles XYZ at each vertexTriangles/Surface Element ~ 150,000Area, Curvature, Thickness, Volume at each vertexThe cortical surface is represented by a finite element model using triangles to cover the cortical surface. The reconstruction is the (mostly automated) process of assigning an xyz to each corner of each triangle. Once the xyz of each corner of each triangle is known, then it is possible to characterize the entire cortical surface in terms of area, distance, curvature, and thickness.
26Cortical Thickness pial surface Distance between white and pial surfacesOne value per vertexmmwhite/gray surface
27Curvature (Radial) Circle tangent to surface at each vertex Curvature measure is 1/radius of circleOne value per vertexSigned (sulcus/gyrus)
28Inter-subject registration Gyrus-to-Gyrus and Sulcus-to-SulcusSome minor folding patterns won’t line upAtlas registration is probabilistic, most variable regions get less weight.Done automatically in recon-allTemplate
29Query Design Estimate Contrast - QDEC Average brain
30Advantages of FreeSurfer Analysis of separate components of volume – thickness and surface areaGeometry is used for inter-subject registration (major sulcal and gyral patterns)2-D surface smoothing versus 3-D volumesmoothing – more biologically meaningful
36Measuring Anisotropy λ1 λ3 λ2 Eigenvectors: the 3 directions Eigenvalues: the rate of diffusion, λ1, λ2 and λ3Apparent diffusion Coefficient (Mean diffusivity)= average of λ1, λ2 and λ3Direction of least resistance to water diffusion, λ1
42Voxel-based Morphometry for dMRI Issues with regular VBM analysisNot-perfect alignmentSmoothing - arbitraryTract-based Spatial StatisticsSmith et al – FMRIBFractional Anisotropy (FA) mapDTI-TK with TBSSHigh level warping using all the tensor information for better alignment
43DTI and SchizophreniaWidespread FA reduction in Schizophrenia versus controls
44DeCC neuroimaging MDD = 153 HC = 153 Matched age and gender Gaussian Process ClassifierLOOCVAccuracy = 59%