BOLD functional MRI Magnetic properties of oxyhemoglobin and deoxyhemoglobin L. Pauling and C. Coryell, PNAS USA 22:210-216 (1936) BOLD effects in vivo S. Ogawa, et al., MRM, 14:68-78 (1990) BOLD activation experiments K. K. Kwong, et al., PNAS USA, 89:5675-5679 (1992) S. Ogawa, et al., PNAS USA, 89:5951-5955 (1992) P. A. Bandettini, et al., MRM, 25:390-397 (1992) J. Frahm, et al., JMRI, 2:501-505 (1992)

Mechanism of BOLD Functional MRI Brain activity Oxygen consumption Cerebral blood flow Oxyhemoglobin Deoxyhemoglobin Magnetic susceptibility T2* MRI signal intesity

Magnetic Properties of Oxyhemoglobin and Deoxyhemoglobin Deoxyhemoglobin: paramagnetic (c > 0) paramagnetic with respect to the surrounding tissue Oxyhemoglobin: diamagnetic (c < 0) isomagnetic with respect to the surrounding tissue

Magnetic Susceptibility

Oxyhemoglobin and Deoxyhemoglobin in Veins during Brain Activation Rest Activation Normal blood flow High blood flow Oxyhemoglobin Deoxyhemoglobin

T2* Effect in fMRI action MR signal (S) rest TE t excitation reception

Time Series and Activation Maps Signal Intensity Off On Off On Off On Off On Scan Number

Challenges in Functional FMRI Sensitivity (Contrast-to-noise ratio) BOLD signal change is ~1-2% at 1.5 T; signal-to-noise ratio in single-shot EPI images is ~100. Physiological pulsations (cardiac and respiratory); Head motion; instrumental instability Specificity Location of activation – neurons or veins Susceptibility artifacts

Challenges in Functional FMRI Temporal resolution Limited by BOLD impulse-response function, image sampling rate, and spin relaxation times Spatial resolution Limited by BOLD point-spread function, signal-to-noise ratio, and image sampling rate Non-linearity Neurological and hemodynamic Acoustic noise

Contrast-to-Noise Ratio Brain Activation-related signal change Sensitivity = Temporal fluctuation of image intensity

Enhancement of BOLD Contrast Higher magnetic fields BOLD signal change DS ~ Ba (1 < a < 2) Standard clinical MRI scanner at 1.5 T Research scanner up to 8 T currently Optimization of image acquisition parameters Optimal echo time (TE) to maximize BOLD signal Optimal repetition time (TR) to increase number of images acquired per unit time, and to decrease motion artifacts

TE Dependence of Signal Change

Suppression of Temporal Fluctuations Head motion reduction Head holder modified from a football helmet Image realignment in data processing Physiological pulsations Correction using simultaneously recorded cardiac and respiratory signals Ultra-fast imaging techniques Single-shot echo-planner imaging (EPI) Single-shot Spiral imaging Post-processing Denoising

Ultra-Fast Spiral Scanning An image (64x64) can be acquired in ~ 20 ms Reduce head motion Increase number of images collected per unit time Stable Spiral trajectory is insensitive to motion and flow artifacts Zero gradient moments at the center of k-space (self-navigated) First-order gradient moments vary smoothly over k-space

Multi-Slice Spiral Images

Multi-Slice EPI Images

Activation Maps on Anatomical Images MS Spiral MS EPI 3D Spiral

Histograms of Temporal Standard Deviations 80 60 40 20 MS-Spiral MS-EPI 3D-Spiral Number of Activated Voxels 0 0.01 0.02 0.03 0.04 SD/Mean

Comparison of Activation Studies Using MS-spiral, MS-EPI, and 3D-spiral

Specificity in fMRI Inflow Effects Generated by fresh (fully recovered) spins moving into the region excited under saturating conditions Flow change in large vessels can lead to substantial signal increases (20-30%), and compromise spatial accuracy in activation studies

Inflow Effects in fMRI Suppression of inflowing spins a b d b Suppression of inflowing spins No suppression of inflowing spins

Gradient Echo vs. Spin Echo in fMRI Diameter (mm) 1.0 10.0 DR2, DR2 * (1/sec) 20 10 DR2 DR2* Contribution of large vessels and capillary beds to the BOLD signals; Separation/suppression of signals from large vessels.

Temporal resolution Impulse-response function 12s 5s 2s t

Temporal resolution Sampling rate (single-shot EPI ~10-15 slices/sec) Whole brain (~30 4mm-slices): 2-3 sec T1 relaxation times Grey matter: 1 sec White matter: 0.8 sec CSF: 2-3 sec

Spatial resolution BOLD point-spread function Image spatial resolution Spatial extent of neuronal activity, CBF, and BOLD Image spatial resolution 64x64 with FOV 240 mm: 3.75mm 128x128 with FOV 240 mm: 1.875 mm Signal-to-noise ration Single-shot EPI with voxel size 4x4x4 mm3: ~100

Non-linearity of BOLD Response BOLD response vs. length of stimulation t 2t BOLD response during rapidly-repeated stimulation ts

Experimental Designs in fMRI Block-Design fMRI Task Rest 20-60s Event-Related fMRI 8-12s

Hemodynamic Response vs. ISI 5 10 15 Time (sec) Signal Changes (%) 1.00 1.01 ISI=8s ISI=12s ISI=16s

Visual Activation Maps (ISI=12s)

Data Analysis Methods For fMRI Hypothesis-driven approaches t-test, cross-correlation, GLM, etc. Data-driven approaches Principal component analysis (PCA), independent component analysis (ICA), and clustering analysis.