Basic MRI physics and Functional MRI Gregory R. Lee, Ph.D Assistant Professor, Department of Radiology June 24, 2013 Pediatric Neuroimaging Research Consortium
Objectives Neuroimaging Overview MR Physics Spins Excitation Relaxation Image Contrast Structural Imaging BOLD fmri
Three Approaches to Functional Neuroimaging Monitor brain activity directly EEG MEG Monitor local cerebral metabolic rate PET SPECT Monitor local CBF and CBV PET fmri Slide Courtesy of Scott Holland
MR Physics: History Nuclear magnetic resonance (NMR) refers to the absorption and re-emission of electromagnetic radiation by nuclei in a magnetic field. Discovered independently by Felix Bloch and Edward Purcell in 1946. (Nobel Prize in 1952). Initially used in the fields of physics and chemistry to study molecular structure (i.e. NMR Spectroscopy) 1973: Paul Lauterbur forms the first MR image using linear gradients. 1980s: MRI developed commercially
MR Physics: Spin Spin is an intrinsic form of angular momentum carried by elementary particles and atomic nuclei. It is a fundamental property of nature such as electrical charge or mass. Particles with spin can possess a magnetic dipole moment, µ. All nuclei with an odd number of protons and/or neutrons have a non-zero magnetic moment. Examples of NMR-active nuclei present in the human body are: 1 H, 13 C, 14 N, 17 O, 19 F, 23 Na, 31 P. 1 H is by far the most abundant and therefore easy to detect.
MR Physics: Spin Think of each hydrogen proton as a tiny bar magnet. 1 gram of your body has ~ 6x10 22 protons. In the absence of an external field, these are randomly oriented and there is no net magnetic moment. In the presence of an external magnetic field, spins can either align with or against the applied field. Image Source: www.physicscentral.com More spins align with the applied field, leading to a net magnetic moment in the sample.
MR Physics: Precession A spin has both magnetization (M) and angular momentum (L): = An applied magnetic field (B 0 ) exerts a force on the magnetization, leading to a torque (T) = = This can be rewritten as the Bloch equation: =
MR Physics: Precession The Bloch equation: = States that the magnetization will precess about the applied magnetic field at a frequency: = This frequency is referred to as the Larmor frequency.
MR Physics: Precession A precessing magnetic moment, will induce an electric current in a MRI receiver coil (via Faraday s Law). The received signal will oscillate at the Larmor frequency. For a 3T magnet, the Larmor frequency for 1 H is approximately 127.6 MHz.
MR Physics: Excitation In MRI, the net magnetic moment is initially aligned with B 0. However, we need to make it perpendicular to B0 so that it will precess and can be detected by our receiver coil. This is done via the application of RF energy at the Larmor frequency in a plane perpendicular to B 0. This RF field is typically applied at a strength of approximately 10µT for a duration of 0.1-10 ms.
MR Physics: Excitation Summary of Basic NMR Experiment: 1.) Place spins in static magnetic field (B 0 ) to create a net magnetic moment. 2.) Apply RF excitation to tip spins into the transverse plane where they will precess. 3.) Acquire MR signal.
MR Physics: Relaxation Following excitation, the spin system will relax back to its equilibrium state. There are two primary components of relaxation: Longitudinal recovery (T 1 recovery) Decay of transverse magnetization (T 2 decay)
MR Physics: T 1 Relaxation T 1 relaxation is often referred to as spin-lattice relaxation and corresponds to the spin giving up energy into the surrounding molecular matrix as heat. T1 determines the rate of recovery of longitudinal magnetization (Mz). Typically 1-3 seconds. =( ) = 0 / + (1 / )
MR Physics: T 1 Relaxation (Recovery)
MR Physics: T 2 Relaxation T 2 relaxation is often referred to as spin-spin relaxation. This term describes the decay of the transverse magnetization (M xy ), with time constant T 2. This decay is due to phase incoherence among spins caused by random field fluctuations caused by intermolecular interactions. In brain tissue T 2 is typically tens of milliseconds. In general, T2 <= T1. = = 0 /
MR Physics: T 2 Relaxation (Decay)
Example with same T2, different ρ ρ= spin density
MR Images of Brain Anatomy T 1 -weighted coronal image T 1,CSF >> T 1 GM > T 1,WM Images Courtesy of Scott Holland T 2 -weighted transverse image T 2,CSF >> T 2 GM> >T 2,WM
MR Physics: T 2* Relaxation T 2* decay is like T 2 decay, but includes the decay from macroscopic static magnetic field inhomogeneity as well. 1 = 1 + 1 Gradient echo images are sensitive to T2* while spin echo images are sensitive to T2.
MRI Hardware Image Source: slide from Tor Wager
MRI Hardware Image Source: slide from Doug Noll
MR Physics: Gradients MR systems are equipped with three sets of linear field gradients: G x, G y, G z. G x : G y : Application of these gradients leads to a linear modulation of the z-component of the B 0 field along either the x, y, or z spatial axis. By the Larmor relation, = ( + + + ) In the presence of a gradient, frequency is position-dependent Image Source: slide from Doug Noll
MR Physics: Gradients An example of two narrow samples placed under a linear gradient field. The Fourier transform is the mathematical tool which allows us to transform a timecourseinto its separate frequency components.
MR Physics: Gradients Gradients are also used for slice selection. RF pulses have a finite bandwidth that can be mapped to a spatial band by use of a gradient pulse during RF excitation:
Conventional Proton MRI yields images of anatomy MRI Signal: ~ (,, ) = proton density = spin-lattice relaxation time = spin-spin relaxation time Differences in,, between tissues produce contrast in conventional MRI.,, are influenced by pathology.
MR Images of Brain Anatomy T 1 -weighted images
MR Images of Brain Anatomy T 2 -weighted images
MR Images of Brain Anatomy CSF GM WM Segmentation Example
fmri = functional Magnetic Resonance Imaging BOLD = blood oxygenation level dependent
How does fmri work? Brain stimulation Local neuronal activity Local increase in blood flow Decrease in deoxyhemoglobin MRI pixel intensity change Slide Courtesy of Scott Holland
Blood Oxygenation Level Dependent MRI Hemoglobin has different magnetic properties depending on whether or not it is oxygenated. The magnetic field in a given tissue is modulated by the magnetic susceptibility, : = 1+ Water, oxygenated blood and other brain tissues have approximately the same magnetic susceptibility. However, deoxygenated hemoglobin has unpaired electrons and is paramagnetic. Deoxygenated blood has a susceptibility of 2.3x10-6 relative to water. (Room air has a relative susceptibility of 9.4x10-6 ).
Blood Oxygenation Level Dependent MRI
Blood Oxygenation Level Dependent MRI As shown in the previous slide, deoxyhemoglobincauses local disturbances in the magnetic field that will lead to signal loss in T2* weighted images. This signal loss will be proportional to the concentration of deoxyhb. The total amount of deoxyhbis proportional to Hct* V * (1-Y) where Hctis hematocrit, V is blood volume, and Y is blood oxygenation.
Blood Oxygenation Level Dependent MRI Image From: KR Thulborn. NeuroImage 62:589-593 (2012)
Blood Oxygenation Level Dependent MRI Souce: Ogawa et. al. PNAS 1992
Slide Courtesy of Scott Holland
Mechanism of BOLD Functional MRI Brain activity Oxygen consumption Cerebral blood flow - + Oxyhemoglobin Deoxyhemoglobin T2* MRI signal intesity Adapted From Slide by of Scott Holland
Blood Oxygenation Level Dependent MRI Brain activation causes local increase in blood flow. Relatively tight coupling between neuronal activity and local CBF, both spatially and temporally. Thompson, J.K., Peterson, M.R., Freeman, RD, Single- Neuron Activity and the Tissue Oxygenation in the Cerebral Cortex. Science, 299:1070-1072, 2003. Single neuron spike rate increase accompanied by immediate decrease in tissue oxygenation suggesting that high-resolution fmri may be used to localize activity in neurons. Slide Courtesy of Scott Holland
Blood Oxygenation Level Dependent MRI Magneticresonance imaging is sensitive to changes in the magnetic properties of blood. BOLD-related signal changes are small (1-5%). Noise in fmri images is typically around 0.5-1% For this reason, we cannot usually detect activation in a single image. Instead, signal changes are detected statistically by time-series analysis of many images. (i.e. we look for regions where the signal time course is highly correlated to the experimental paradigm) Slide Courtesy of Scott Holland
Hemodynamic Response to Brain Activity BOLD effect is not instantaneous Initial dip or undershoot of BOLD signal ~ 1 s Due to lag in CBF relative to O 2 demand Delay between activity and BOLD response ~ 4-6 s Decay of BOLD signal after activity stops ~ 6-8 s BOLD fmri is good for measuring sustained localized cortical activity but too slow to get dynamic information. Slide Courtesy of Scott Holland
Hemodynamic Response Function (HRF) Response to Single Short Stimulus Time (sec) Slide Courtesy of Scott Holland
Statistic Post-Processing BOLD signal is small 1-2% at 1.5T 2-4% at 3.0T Statistical methods are needed To detect activation above noise To determine significance of activation To assign probabilities to activation Statistical parameter mapping T-test Cross-correlation methods General linear model ICA/PCA Slide Courtesy of Scott Holland
Block Periodic fmri Paradigms Alternating blocks of behavior Stimulus (30 seconds typical) Control (30 seconds typical) Complete paradigm contains multiple blocks 5 x {Stimulation/Control} Increases Statistical Power Total acquisition time per task = 5 min. 30 sec. Control task minimizes incidental activation Control for sensory stimuli Control for attention Distract from target behavior Slide Courtesy of Scott Holland
Blood Oxygenation Level Dependent MRI A BOLD MRI acquisition is typically performed as a series of 2D slices which cover the whole brain. TE ~=T2* to optimize BOLD contrast. Longer TE -> greater signal loss near air-tissue interfaces. Echo-Planer Imaging (EPI) allows acquisition at a rate of approximately 50 ms/ slice. Whole brain in ~1-2 s. Typically use 3-5 mm voxels to enable single-shot EPI and to have adequate SNR. skh.111398.qc1.no6
Pixel Intensity Time Course with Periodic Stimulation and Rest REST 1 ACTIVE 1 REST 2 ACTIVE 2 REST 3 30 sec 30 sec 30 sec 30 sec 30 sec random short random short random tones story tones story tones Stimulus Stimulus Stimulus Rest Rest Rest Rest time Slide Courtesy of Scott Holland
fmri of Passive Story Listening Task 35 y.o. Female Listening to Stories @ 3T 102 Normalized Activation 101 100 99 skh.111398.qc1.no6 Slide Courtesy of Scott Holland 98 0 30 60 90 120 150 180 210 240 270 300 Time (sec) Time course in temporo-parietal language cortex, Wernicke s area.