What is Blood Oxygenation Level Dependent (BOLD) signal

Transcription

1 What is Blood Oxygenation Level Dependent (BOLD) signal Simo Vanni Brain Research Unit, O.V. Lounasmaa Laboratory Advanced Magnetic Imaging Centre Aalto University School of Science

2 Internet site for the fmri school

3 Spatial resolution Resolution of brain imaging methods dm EEG cm MEG PET fmri mm ms s min Temporal resolution

4 Spatial resolution Resolution of brain imaging methods dm EEG cm MEG PET mm fmri ms s min Temporal resolution

5 Classical activation study: pain network activation in empathy Singer et al. Science 303 (2004)

6 Human visually responsive areas mapped with fmri Vanni, Duodecim 120 (2004)

7 Human ocular dominance columns -7 T magnet (Varian), 0.5 x 0.5 x 3 mm voxels Day 1 Day 2 Day 3 -Hahn Spin Echo runs á 7 min (2-2.5 h/ session) -bite-bar -post-processing 1 mm Yacoub et al. Neuroimage 37 (2007)

8 How does fmri create images of neuronal activity? It creates images of changes in deoxyhemoglobin concentration correlated with neuronal activity Blood Oxygenation Level Dependent contrast (BOLD) deoxygenated blood paramagnetic oxygenated blood diamagnetic

9 BOLD contrast generation Gradient-echo MR image Oxygenated blood Saline Blood in test tube corrupted MR signal Deoxygenated blood From Huettel, Song, McCarthy (eds) Functional magnetic resonance imaging. Sinauer 2004

10 Blood oxygen level dependent contrast 100% inhaled oxygen (a) shows veins as dark stripes 90% oxygen with 10% carbon dioxide (b) show decreased BOLD contrast, because blood flow has increased Ogawa et al. PNAS 87 (1990)

11 BOLD signal BOLD contrast overperfusion initial dip undershoot Time From Huettel, Song, McCarthy (eds) Functional magnetic resonance imaging. Sinauer 2004

12 Neurovascular coupling

13 BOLD reflects synaptic activity Visual cortex of monkeys Comparison of synaptic activity (LFP), action potentials (MUA) And BOLD Logothetis et al. Nature 412 (2001)

14 BOLD reflects synaptic activity Logothetis et al. Nature 412 (2001)

15 From blood flow to BOLD From Huettel, Song, McCarthy (eds) Functional magnetic resonance imaging. Sinauer 2004

16 Astrocytes and neurovascular coupling Human astrocyte Contact hundreds of neuronal dendrites Sense > millon synapses At least one process with endfeet surrounding blood vessels Oberheim et al. TINS 29 (2006) ; Koehler et al. TINS 32 (2009)

17 Astrocytes and neurovascular coupling Neural activation Astrocyte Vasodilation Koehler et al. TINS 32 (2009)

18 Coupling of synaptic activation and BOLD signal Astrocytes key players in neurovascular coupling Tight regulation Multiple pathways Blood flow BOLD connection is a monotonous function Blood volume changes cause compressive nonlinearity

19 Why excessive perfusion? Increased oxygen gradient between blood vessels and tissue Adequate oxyden diffusion to most distant neuron (mitochondria) Koehler et al. TINS 32 (2009)

20 Measuring BOLD signal

21 Canonical hemodynamic response function (hrf) Event, impulse response Epoch, block

22 Functional magnetic resonance imaging (fmri) is a time series of MR images one slice image matrix, e.g. 64 x 64

23 fmri time series

24 Definition of voxel dimensions Matrix size (N x N) Field of view (cm) Slice thickness (mm)

25 Latency between successive time points is Repetition Time (TR) - time point - volume - stack

26 Partial volume effect A single voxel contains a mixture of different tissues White matter Gray matter

27 Partial volume effect A single voxel contains a mixture of different tissues Brain Cerebrospinal fluid

28 BOLD signal BOLD signal Time (time points)

29 BOLD signal BOLD signal Time (time points)

30 BOLD signal

31 Canonical hrf is not perfect model Handwerker et al., Neuroimage 21 (2004)

32 Multi-channel coils and parallel imaging

33 Physiological noise dominates at high SNR Triantafyllou et al., Neuroimage 55 (2011)

34 Spatial resolution of BOLD signal

35 Distribution of arterioles 1 Cortical arterioles 2 Pial arterias 3 Precapillary vessels Scale bar = 190 μm Rodriguez-Baeza et al. Anat. Record 252 (1998),

36 Distribution of arterial vessels Temporal Occipital Reina-De La Torre et al. Anat. Record 251 (1998), 87-96

37 Resolution of fmri Ultimate theoretical spatial resolution gradient-echo imaging 0.7 mm 3 Spin echo imaging 0.44 mm 3 (arterial feeding volume) Resolution in practice more limited Weber et al. Cerebral Cortex 18 (2008)

38 The simple reasons for loss of spatial resolution and specificity Motion during acquisition or between runs Poor coregistration with anatomical image Smoothing in preprocessing Normalization between subjects Region of interest analysis Partial volume effects

39 What limits the resolution in practice 1) Signal to noise -ratio (SNR) Voxel volume ~ MR signal / unit time Careful design and pilot studies necessary 2) Magnetic field inhomogeneity Distorts the EPI volume, especially close to air cavities Dependent on acquisition time (long in EPI imaging) Shimming helps, automatic Magn. field mapping and distortion correction possible

40 Anisotropic and patchy horizontal connections ~3 mm Anisotropy follows cortical magnification in V1 0.5 mm 0.1 mm Double labelled cells indicate resiprocal connections Angelucci et al., J Neuroscience 22 (2002)

41 Neural point spread in monkey cortex Grinvald et al., J Neuroscience 14 (1994)

42 All direct quantitative comparisons should be made within single voxels in balanced designs BOLD signal strength varies accross runs, brain areas and subjects