Perfusion-Based fMRI Thomas T Perfusion-Based fMRI Thomas T. Liu Center for Functional MRI University of California San Diego May 7, 2006

Goal To provide a basic understanding of the theory and application of arterial spin labeling in functional MRI.

Topics What is Cerebral Blood Flow? Arterial Spin Labeling Data Processing Applications

Cerebral blood flow (CBF) is a fundamental physiological quantity Cerebral blood flow (CBF) is a fundamental physiological quantity. Closely related to brain function. From C. Iadecola 2004

Cerebral Blood Flow (CBF) CBF = Perfusion = Rate of delivery of arterial blood to a capillary bed in tissue. Units: (ml of Blood) (100 grams of tissue)(minute) Typical value is 60 ml/(100g-min) or 60 ml/(100 ml-min) = 0.01 s-1, assuming average density of brain equals 1 gm/ml

High CBF Low CBF Time

Bereczki et al 1992

Methods for Measuring CBF Quantitative Autoradiography Laser Doppler Diffusible and Radioactive Tracer Based Methods PET and SPECT Contrast-Based MRI Arterial Spin Labeling

Arterial spin labeling (ASL) Tag by Magnetic Inversion 1: Acquire image Control 2: Acquire image The technique of arterial spin labeleing can be defined as follows. We first tag arterial blood prior to the imaging plan we then wait and aquire an image. Control - Tag = M µ CBF

Flow-driven inversion Velocity-selective ASL Imaging Region Pulsed ASL Spatial inversion Tag Continuous ASL Flow-driven inversion Velocity-selective ASL Velocity-dependent saturation Mz(blood) M control tag t TI Mz(blood) M control tag t TI

ASL Pulse Sequences Tag Control Acquire TI = Inversion Time TR PASL/VSASL Tag Control Acquire Labeling Time Post Labeling Delay CASL

Conventional Pulsed ASL (PASL) off-resonance IR pulse tag imaging slice control presaturation slice off-resonance IR pulse EPISTAR PICORE FAIR Courtesy of Wen-Ming Luh

Continuous ASL (CASL) B0 Amplitude Conventional Tag Modulated Control Imaging Plane Blood Magnetization Inversion Planes B0 Amplitude Modulated Control Conventional Control Tag

Velocity Selective ASL (VSASL) 0.1 Velocity (cm/s) Physiological Motion 1 10 Mz { Control Tag Image Wong et al 2006

CASL PASL PASL VSASL Credit: E. Wong

ASL Signal Equation ∆M= CBF · Aeff Mz(blood) control M - = M tag t t TI ∆M= CBF · Aeff Aeff is the effective area of the arterial bolus. It depends on both physiology and pulse sequence parameters.

Quantitative ASL Goal: Make Aeff a well-controlled parameter ∆M= CBF · Aeff Goal: Make Aeff a well-controlled parameter that is robust to assumptions about physiological parameters.

Major Sources of Error for ASL Transit Delays Bolus Width in PASL Relaxation Effects - different relaxation rates for blood and tissue, time of exchange. Intravascular signal -- blood destined to perfuse more distal slices Offset bias to due imperfect subtraction of static tissue -- slice profiles, magnetization transfer effects

Transit Delays CASL PASL ~ 3 cm ∆t < 1000ms ~ 1 cm ∆t < 700ms

Controlling for Transit Delays in CASL Tagging Plane A B A B time Voxels A and B have the same CBF, but voxel B will appear to have lower CBF if the measurement is made too early. Proper selection of post-labeling delay is key

Amplifying Transit Delays Effects in CASL Tagging Plane Baseline Signal Activation Signal Baseline Signal Activation Signal time Acquiring the signal at an earlier TI amplifies the difference between the activated state and the baseline state.

Arterial Bolus Width CASL PASL time Temporal Width of Baseline bolus determined by the pulse sequence Baseline Global flow increase PASL Temporal Width of bolus determined by arterial velocity and size of tagging slab. Underestimates global flow changes. Baseline Global flow increase time

Defining Bolus Width in PASL (QUIPSS II) Saturate spins still in the slab Baseline Global flow increase Tag the spins TI1 Bolus temporal width = TI1

QUIPSS II Modification Tag Control Acquire TI1 Saturation Pulses

Controlling for Transit Delays in PASL Tagging Slab A B A B TI1 TI2 > transit delay + TI1

Topics What is Cerebral Blood Flow? Arterial Spin Labeling Data Processing Applications

ASL Time Series Wait 1 +0.5 -0.5 -1 Perfusion Images Image 1 Image 2 Tag by Magnetic Inversion Tag Image 1 Control Image 2 Tag by Magnetic Inversion Tag Image 3 Control Image 4 -0.5 1 +0.5 -1 Perfusion Images

CBF and BOLD Time Series

Demodulate Modulate

Modulate

Hernandez-Garcia et al 2005; Roller et al 2006, Abstract 540

Removing Physiological Noise Modeling a fMRI signal using a General Linear Model (GLM) + = Data Physiological Noise Matrix Additive Gaussian Noise Nuisance Matrix Design Matrix To estimate the contribution of physiological noise to the fMRI signal, we can model our fmri signal as a General linear model To estimate the contribution of X, S, and P to y, h, b, and c are estimated using least squares estimation. RETROICOR assumes the physiological noise is additive, thus estimating c allows us to subtract Pc from y to get a noiseless image time series. H-x*xinvxxy In the following slides I will dexcribe how we model matrices X S and P y = Xh + Sb + Pc + n

Removing Physiological Noise Perfusion time series: Before Correction (cc = 0.15) Perfusion time series: After Correction (cc = 0.71) Perfusion Time Series Signal Intensity Image Number (4 Hz) In our methods, we have treated the Tag/Control time series as the signal of interest, but it is useful to see how our method corrects a perfusion time series since that is the signal that we care about. Pulse Ox (finger) Cardiac component estimated in brain

Removing Physiological Noise 0.200 0.275 0.350 0.425 0.500 Young Subject Older Subject (b) Corrected (a) Uncorrected Corrected (c) (d) (e) In our methods, we have treated the Tag/Control time series as the signal of interest, but it is useful to see how our method corrects a perfusion time series since that is the signal that we care about.

Principal Components Based Correction mRETROICOR CompCor Behzadi et al Abstract 235

Background Suppression Tag Acquire QUIPSS II Saturation Pulse Inversion Pulses Pre-saturation Pulse 200 400 600 800 1000 1200 1400 -1 -0.8 -0.6 -0.4 -0.2 0.2 0.4 0.6 0.8 1 time (ms) M z Background Suppression with Double Inversion GM WM CSF Nulling of static tissue element

Background Suppression Visual Study Results St. Lawrence et al 2005 Behzadi et al Abstract 3295

Single Shot Perfusion Labeling (SSPL) From JA Duyn et al, MRM 2001

Single Shot Perfusion Labeling (SSPL) From JA Duyn et al, MRM 2001

Topics What is Cerebral Blood Flow? Arterial Spin Labeling Data Processing Applications

Applications of ASL Quantitative Measures of both baseline and functional CBF Multimodal measures of CBF, BOLD, (and CBV) to estimate functional changes in oxygen metabolism. Experiments with long task periods Mapping of functional activity.

Baseline CBF Modulates the BOLD Response

Effect of Baseline CBF on BOLD Response

Effect of Acetazolamide on BOLD and CBF Percent BOLD Changes CBF Changes Brown et al 2003

Effect of Hypercapnia on BOLD and CBF Stefanovic et al 2006

Applications of ASL Quantitative Measures of both baseline and functional CBF Multimodal measures of CBF, BOLD, (and CBV) to estimate functional changes in oxygen metabolism. Experiments with long task periods Mapping of functional activity.

BOLD Dynamics CMRO2 Stimulus ATP ADP BOLD Neural CBF Signal Activity CMRGlc CBF CMRO2 CBV BOLD Signal NO, K+ CO2 Lac Credit: Rick Buxton

Effect of Age on CBF and BOLD Responses in the Hippocampus Restom et al, Abstract 377

Estimation of CMRO2 Changes with Combined CBF and BOLD measures Hypercapnia Contralateral Ipsilateral Stefanovic et al 2004

Simultaneous measure of CBF, BOLD, and VASO (CBV) Yang et al 2004 VASO CBF BOLD

Bayesian Analysis of dual -echo ASL data (Woolrich et al. Abstract 538) TE=9 msecs TE=30 msecs data model fit static delivered blood

Bayesian Analysis of ASL data (Woolrich et al. Abstract 538) CBF static magnetisation, M

Applications of ASL Quantitative Measures of both baseline and functional CBF Multimodal measures of CBF, BOLD, (and CBV) to estimate functional changes in oxygen metabolism. Experiments with long task periods Mapping of functional activity.

ASL with very low task frequencies - Wang et al., MRM 2003

ASL Imaging of Stress- Wang et al., PNAS 2005

ASL Imaging of Natural Vision; Rao et al, Abstract 167 CBF and BOLD activations during viewing of Road Runner “Gee W-h-I-z-z-z-z-z-z” Cartoon

Applications of ASL Quantitative Measures of both baseline and functional CBF Multimodal measures of CBF, BOLD, (and CBV) to estimate functional changes in oxygen metabolism. Experiments with long task periods Mapping of functional activity.

ASL Mapping of Cortical Columns in Cat Visual Cortex Duong et al, PNAS, 2001. FAIR sequence, TI = 1500 ms, TR 3000 ms

Z. Liu and X. Hu, Abstract 891

Summary Arterial spin labeling can provide non-invasive quantitative measures of baseline and functional CBF. 2. Since CBF is a well-defined physiological quantity, measures of CBF can be used to improve the interpretation of fMRI studies, especially for clinical populations. 3. The combination of CBF measures with BOLD (and CBV) measures can be used to estimate changes in oxygen metabolism.

Acknowledgements Yashar Behzadi Rick Buxton Wen-Ming Luh Joanna Perthen Khaled Restom Eric Roller Mark Woolrich Eric Wong