fmri: Interpretation, Limits and Potential Pitfalls Seong-Gi Kim kimsg@pitt.edu www.kimlab.pitt.edu
Mapping Brain Functions Stimulation/Task Functional Map (MRI) Pre-synaptic activity Post-synaptic activity Action potentials Neural Activity Blood flow Blood volume Blood oxygenation Vascular Response
Vascular Structure Arteries Capillaries Veins Blood oxygenation level ~1.0 ~0.6 Distance
Blood Oxygenation Level-dependent Contrast Ogawa et al. Magn. Reson. Med, 1990 HEMOGLOBIN Oxyhemoglobin (Diamagnetic) -O 2 Deoxyhemoglobin (Paramagnetic) Reduce T 2 * -> Reduce signal intensity in T 2 *-weighted images
T 2 *-weighted images of rat brain (no activation) (isotropic resolution of 58 μm, 9.4 T) 4mm Dark lines venous vessels (>20 micrometer diameters) Sung-Hong Park et al., Magn. Reson. Med., 2008
Blood Oxygenation Level-dependent Contrast Ogawa et al. MRM, 1990 Breathing air Breathing 100% O 2 Mouse brain images at 360 MHz
Dynamic BOLD MR Measurements in Cats Turner R, Le Bihan D, Moonen CT, Despres D, Frank J Echo-planar time course MRI of cat brain oxygenation changes Magn Reson Med. 1991 Nov;22(1):159-66 Abstract: When deoxygenated, blood behaves as an effective susceptibility contrast agent. Changes in brain oxygenation can be monitored using gradient-echo echo-planar imaging. With this technique, difference images also demonstrate that blood oxygenation is increased during periods of recovery from respiratory challenge.
Vascular Structure Arteries Capillaries Veins Blood oxygenation level ~1.0 (Task -> oxygen supply overcompensates oxygen utilization) ~0.6 Distance (Fox et al., 1988)
One of First Human fmri Studies Primary Visual Cortex Anatomical Image Functional Image (Visual Stimulation) University of Minnesota/Bell Lab Ogawa et al. Proc Natl Acad Sci USA, 1992
Current Status of Functional MRI - Underlying assumption is that fmri signal change is indirectly related to neural activity, and its location is indicative of neural activity site. - Functional MRI with a few millimeter resolution is routinely used for mapping brain functions such as vision, motor, language, cognition, etc.
Physiological Changes Biophysical Basis of BOLD fmri Spatial Resolution Interpretation - Quantification Temporal Resolution
Vascular Physiological Changes Blood Vessel Dilation Blood Velocity Increase Cerebral Blood Flow Blood Oxygenation Change Costantino Iadecola & Maiken Nedergaard Nature Neurosci, 2007
200µm Vessel Imaging of Rat Brain Anatomical Image 1 mm Vazquez et al., High-resolution Anatomical Image 500 µm
Vessel Imaging of Rat Brain during Stimulation 200µm 20x Mag., Reverse contrast A Bright: dilation 1 mm V V 4-s forepaw stim Vazquez et al.
Simultaneous measurements of CBF and P O2 QUANTIFICATION (PO2) Venous PO2 LV MV T SV SA MA Arteries Veins Tissue PO2 LDF (CBF) LA Clark-type oxygen sensor (30 and 4 μm diameter) Vazquez et al., JCBFM, 2010
CBF and tissue PO2 changes during stimulation Forepaw Stimulation LDF (CBF) Tissue PO2 Time (s) Vazquez et al., JCBFM, 2010
Venous Blood PO2 changes during stimulation SO2 Sm. Ven. Med. Ven. PO2 Lar. Ven. Vazquez et al., JCBFM, 2010
Physiological Changes Biophysical Basis of BOLD fmri Spatial Resolution Interpretation - Quantification Temporal Resolution
Compartmentalization of Water Intravascular water moves freely EV blood vessel (IV) Slow exchange of IV and EV water Intact BBB tight junctions between endothelial cells impede the diffusion of water. (In 50 ms, less than 5% of the capillary water diffuses into the EVS.) Extravascular water moves freely RBC
Intravascular Effect -> T 2 Change Reb Blood Cell water Water appears to move freely across the RBC membrane (residence time in RBC ~ 5 ms).
Susceptibility effect in Extravascular Pool: Δω out = Δω max (radius of vessel/distance from vessel) 2 Related to deoxyhemoglobin concentration (oxygen saturation level) magnetic field strength 30 μm 300 μm 1% of max at r = 10 a.
Susceptibility effect in Extravascular Pool: ΔBout MRI signal at echo time (TE): a summation of all water proton signal within a voxel. Each proton signal decays by T 2 and dephases by local susceptibility effect (i.e., Phase shift) S(TE) = S. exp(-te/t 2 ). e(-iϖte) S. exp(-te/t 2 *) 30 μm 300 μm 1% of max at r = 10 a.
Spin Echo (two spins) t = 0 (after 90 pulse) x τ x y y 180 pulse along x Spin-echo x τ x y y
Capillary tube (1.4 mm o.d., 1.0 mm i.d.) filled with blood in a saline bath (positioned orthogonal to main magnetic field) SE GE 100% oxyhb 100% deoxyhb Ogawa et al., MRM, 1990
Conventional Gradient-echo and Spin-echo BOLD Signal CBV = 2% Δχ = 0.1 ppm Boxerman et al., MRM, 1995
Extravascular and Intravascular BOLD Signal Contributions Gradient Echo Spin Echo Extravascular Large vessels Small vessels X X X Intravascular Large Small X X X X
Physiological Changes Biophysical Basis of BOLD fmri Spatial Resolution Interpretation - Quantification Temporal Resolution
Since all fmri techniques rely on blood signals, it is desirable to detect responses of small vessels which are close to active neurons. Midline Human visual cortex ~ 2 mm White matter Vascular Structures - Histology Duvernoy et al. Brain Research Bulletin, 1981
Cortical Layer Model Layer 4 is known to be highest capillary density and metabolic responses. Pia Matter Pia 1 2 1 2 3 4 5 6 D Gray Matter 3 4 WM 2 mm L 5 White Matter Vascular structure 6 white Cortical cytoarchitecture of cat visual area 18 -Timan et al., Brain Res, 2004 - Torre et al., Anat Rec 1998
Cortical Depth-Dependent Gradient-echo BOLD fmri (156 x 156 µm 2 in-resolution, 4-shot EPI, 9.4T) % Change (TE=20 ms, 9.4T) 6 5 GM WM 4 3 2 1 0-0.5 0 0.5 1 1.5 2 2.5 Distance from Cortical Surface (mm) 2 mm Zhao et al., NeuroImage, 2006
Vascular structures vs. fmri resolution Scanning Electron Microscopy Pia (Human cortex) mater Torre et al., 1998 Gray matter BOLD Signal 500 µm Δdeoxyhemoglobin conc. in blood x venous blood volume
Gradient-echo vs. Spin-echo BOLD fmri (156 x 156 µm 2 in-resolution, 4-shot EPI, 9.4T) Gradient-Echo Spin-Echo TE=20 ms TE=40 ms 2 mm 0.3 3.0 (%) 0.3 3.0 (%)
Spatial Specificity of BOLD Signal to Neural Activity Site - Venous Vascular Structures Pial Venous Vessels: 130 380 μm diameter Intracortical Veins: 80 120 μm average diameter 1 2 mm apart
Distance between Intracortical Veins artery vein Pia GM WM Distance between emerging venous veins: 0.75-4 mm Duvernoy et al. Brain Research Bulletin, 1981
Can you map cortical columns? Neurons with similar properties are clustered as columns Single-neuron Activities Ocular Dominance Columns Color sensitive regions Gray matter (1.5 3 mm) Orientation columns Hubel & Wiesel, 1968
Iso-orientation maps in the medial area using fmri (with contrast agent, dilation of small vessels) 5 mm D A P V SPL -10 0 +10 Signal intensity (arbitrary unit) Fukuda et al., J of Neurosci, 2006
Observation of orientation preference maps 5 mm 180 A D V P SPL 90 0 Fukuda et al., J of Neurosci, 2006
Left Right Coronal plane mg Marginal gyrus (mg) LS WM Lateral sulcus (LS) BOLD fmri 0.8 0 o 90 o 5 mm 0.3 1mm (Kim et al. Nature Neurosci, 3: 164-169, 2000)
BOLD vs. CBV is-orientation maps (obtained with the differential approach; 0 90 ) GE BOLD SE BOLD CBV 1.0 A -1.0 ΔS (x mean) R 5 mm GE SE CBV 2 mm Moon et al., J of Neurosci, 2007
Physiological Basis Biophysical Basis Spatial Resolution Interpretation - Quantification Temporal Resolution
BOLD Signals Dependent on Bo, TE, pulse sequence (GE vs SE) Dependent on vessel size, orientation, and density Dependent on hematocrit level Dependent on oxygenation level
ΔR 2 * = Δ(1/T 2 *) = percent change/te CBV v (1 ΔS v ) + ΔCBV v (1 S v ) where 1 - S v = CMRO 2 / CBF Cerebral Oxygen Consumption Rate Cerebral Blood Flow Venous Blood Volume
Parenchymal Microvessel (<50 μm diameter) Region Blood volume Occipital cortex 1% Corpus callosum 0.4% Cerebellar nuclei 1.3% Rat; Fenstermacher et al.
Task/stimulation Neural activity CBF CMRo 2 CBV v dhb T 2 * T 2 * BOLD Signal BOLD Signal
1.6 CBV vs. CBF during Hypercapnia (α-chloralose anesthetized rats) rcbv (arbitrary unit) 1.4 1.2 1.0 0.8 0.6 58 ml/100 g/min rcbv = 0.975rCBF 0.40 rcbv = 0.31 rcbf + 0.67 ( r = 0.85 ) (100% CBF -> 31% CBV) 0.4 1.0 1.4 1.8 2.2 rcbf (arbitrary unit) Lee et al., MRM, 2001
rcbv (arbitrary unit) 1.8 1.6 1.4 1.2 1.0 0.8 CBF vs. Arterial and Venous CBV rcbv (vein) rcbv (artery) 0.6 0.6 0.8 1.0 1.2 1.4 1.6 1.8 rcbf (arbitrary unit) Lee et al., MRM, 2001
Task/stimulation Neural activity CBF CMRo 2 CBV dhb T 2 * T 2 *
fmri Signal Change is related to Neural Activity LOGOTHETIS et al. Nature, 412, 150 157, 2001
fmri Signal Change is related to Neural Activity LOGOTHETIS et al. Nature, 412, 150 157, 2001
Visual Stimulation under different baseline conditions Normalized BOLD Signal 1.09 1.06 1.03 1 0.97 hypercapnia 55 50 45 40 35 30 ETCO2 (m m Hg) 0.94 0 100 200 300 400 Time (seconds) n = 6 subjects for each study 25 hypocapnia Cohen et al. JCBFM, 2002
Average BOLD Change (%) 6 4 2 0 Average BOLD Change (%) 1.2 0.8 0.4 0-0.4 0 0.5 1 1.5 2 2.5 3 Time (seconds) hypocapnia normocapnia hypercapnia -2-4 0 4 8 12 16 20 24 28 32 36 40 Time (seconds)
Interpretation of fmri signals - fmri signal is an index of ensemble of neural activity (presumably monotonic relation) - Neural source of BOLD signal is not clear spiking activities vs. synaptic activity, excitatory vs. inhibitory - Difficulty to compare fmri signals across cortical regions and subjects due to BOLD signal dependencies on vascular structure and volume. - Excellent non-invasive tool to map whole brain functions with relatively high spatial (a few millimeters in humans) and temporal resolution (~a few seconds).
Physiological Basis Biophysical Basis Spatial Resolution Interpretation - Quantification Temporal Resolution
Heterogeneity of fmri changes in humans: response times (Bilateral finger movements) Relative Delay Time + 2 sec 0 sec - 2 sec 105 104 103 102 101 100 # of pixels 50 40 30 20 10 99 0 2 4 6 8 10 12 14 16 18 20 Time (sec) 0-2 -1.5-1 -0.5 0 0.5 1 1.5 2 Relative Delay (sec) Provided by P.A. Bandettini
Task Execution Task Execution 2 sec BOLD response BOLD response Time to peak Inter-epoch delay time (1 10 sec)
fmri Signal vs. Finger Movements BOLD change (%) finger pressure 25 20 15 10 5 0-5 white matter motor area Delay time= 3 7.5 s 0 20 40 60 80 100 time (s)
mental rotation experiment displayed until decision is made time Presentation Contemplation Decision Richter et al. J. Cogn. Neurosci, 1999
Functional Maps of Mental Rotation Supplementary Motor Area Central Sulcu Lateral Premotor Area Superior Parietal Area
Response Time-locked Time Courses in M1 and SMA Relative fmri intensity 1% Primary motor Supplementary motor -10-5 0 5 10 15 Time from button press (sec)