7.2 Magnetic Resonance Imaging (MRI)

MRI Principles and Physics

Nuclear Magnetic Resonance and Spin Physics

MRI generates images by exploiting nuclear magnetic resonance (NMR), the interaction between atomic nuclei's magnetic properties and external magnetic fields. The key nucleus in clinical MRI is hydrogen ($^1H$), which is abundant in water and fat throughout the body.

Hydrogen protons have a positive charge and intrinsic spin, which together create a small magnetic field called a magnetic moment. Normally, these spins point in random directions, so the net magnetization across a tissue volume is zero.

When you place a patient inside a strong external magnetic field ($B_0$), protons align either parallel (low-energy) or antiparallel (high-energy) to the field. A slight excess of protons settle into the parallel state, producing a net magnetization vector (NMV) along $B_0$.

These aligned protons don't just sit still. They wobble (precess) around the $B_0$ axis at a characteristic rate called the Larmor frequency, defined by:

$\omega_0 = \gamma \cdot B_0$

where $\gamma$ is the gyromagnetic ratio of the nucleus. For hydrogen, $\gamma = 42.58 \text{ MHz/T}$. So in a 1.5T scanner, protons precess at about 63.87 MHz. This frequency dependence is what makes spatial encoding possible.

Radiofrequency Pulses and Relaxation

To generate a signal, you need to disturb the equilibrium magnetization. An RF pulse tuned to the Larmor frequency transfers energy to the protons (resonance), tipping the NMV away from $B_0$ and creating a measurable transverse magnetization component. The flip angle depends on the pulse's duration and amplitude. Common flip angles are 90° (tips NMV fully into the transverse plane) and 180° (inverts the NMV).

Once the RF pulse switches off, the system returns to equilibrium through two simultaneous but independent relaxation processes:

• T1 relaxation (spin-lattice): The longitudinal magnetization recovers as protons release energy to the surrounding molecular lattice. The time constant T1 varies by tissue. For example, at 1.5T, cerebrospinal fluid (CSF) has a T1 of ~4000 ms, gray matter ~900 ms, and fat ~250 ms. Tissues with efficient energy transfer to their environment recover faster (shorter T1).
• T2 relaxation (spin-spin): The transverse magnetization decays as protons lose phase coherence due to interactions with neighboring spins. T2 is always shorter than or equal to T1 for a given tissue. At 1.5T, muscle has a T2 of ~35 ms, while CSF is ~2000 ms. In practice, local field inhomogeneities cause even faster dephasing, characterized by $T2^*$ (T2-star), which is always shorter than T2.

The RF signals emitted during relaxation are what the scanner detects and uses to build images.

MRI System Components

Main Magnet and Gradient Coils

The main magnet produces the strong, uniform $B_0$ field that polarizes proton spins. Most clinical systems use superconducting magnets cooled by liquid helium, with field strengths of 1.5T or 3T being standard. Research systems can reach 7T or higher. The field must be extremely homogeneous across the imaging volume for accurate results.

Gradient coils create controlled, linear variations in the magnetic field along the x, y, and z axes. These small field variations are what allow spatial encoding of the MRI signal. Each axis serves a specific role in a standard imaging sequence:

• Slice selection ($G_z$): A gradient along one axis ensures that only protons in a specific slice have the correct Larmor frequency to respond to the RF pulse.
• Frequency encoding ($G_x$): Applied during signal readout, this gradient makes protons at different positions along one in-plane axis precess at slightly different frequencies.
• Phase encoding ($G_y$): Applied briefly before readout, this gradient imparts position-dependent phase shifts along the other in-plane axis.

Typical gradient strengths range from 20 to 80 mT/m. Stronger, faster-switching gradients enable higher resolution and faster imaging.

RF Coils and Shim System

RF coils serve two functions: transmitting RF pulses to excite protons and receiving the emitted signals during relaxation. Transmit coils generate a uniform $B_1$ field perpendicular to $B_0$. Receive coils are optimized for signal-to-noise ratio (SNR) and are often designed for specific body regions (head coil, knee coil, cardiac array, etc.). Many modern systems use the body coil for transmission and dedicated surface coil arrays for reception.

Shim coils correct for inhomogeneities in the $B_0$ field, which is critical for image quality. Two approaches are used:

• Passive shimming: Ferromagnetic materials (e.g., iron shim plates) are placed at specific locations inside the magnet bore during installation.
• Active shimming: Additional electromagnetic coils generate small corrective fields that counteract remaining inhomogeneities, and these can be adjusted for each patient.

Patient Table and Computer System

The patient table positions the patient within the magnet bore and must be constructed entirely from non-magnetic materials. It slides in and out of the scanner and supports a range of patient sizes.

The computer system orchestrates the entire scan. It controls pulse sequence timing, coordinates gradient and RF activity, and reconstructs images from the raw data (stored in k-space). The system includes an operator console, a reconstruction engine, and a PACS-compatible storage/archiving system.

MRI Pulse Sequences and Applications

Spin Echo and Gradient Echo Sequences

Spin echo (SE) sequences use a 90° excitation pulse followed by one or more 180° refocusing pulses to generate an echo. The 180° pulse corrects for dephasing caused by static field inhomogeneities, which means SE sequences produce true T2 contrast rather than T2* contrast. This makes them less sensitive to susceptibility artifacts and excellent for anatomical imaging and pathology detection. By adjusting TR (repetition time) and TE (echo time), you can produce T1-weighted, T2-weighted, or proton density-weighted images.

Gradient echo (GRE) sequences use a variable flip angle (often less than 90°) and gradient reversals instead of 180° refocusing pulses to form an echo. Because there's no refocusing pulse, GRE sequences are sensitive to $T2^*$ effects and magnetic susceptibility differences. They're faster than SE sequences and are commonly used for:

• Time-of-flight (TOF) angiography
• Dynamic susceptibility contrast (DSC) perfusion imaging
• Functional MRI (fMRI)
• Rapid anatomical surveys

Inversion Recovery and Echo Planar Imaging

Inversion recovery (IR) sequences add a 180° inversion pulse before the standard excitation pulse. By choosing the right inversion time (TI), you can null the signal from a specific tissue, dramatically improving contrast.

• FLAIR (Fluid-Attenuated Inversion Recovery): Suppresses CSF signal, making periventricular and cortical brain lesions much easier to see (e.g., multiple sclerosis plaques, small infarcts).
• STIR (Short Tau Inversion Recovery): Suppresses fat signal, which is useful for detecting pathology in fatty tissues (e.g., bone marrow edema, soft tissue inflammation).

Echo planar imaging (EPI) acquires an entire image (or most of one) after a single RF excitation by rapidly switching gradients to collect multiple echoes within one TR. This speed makes EPI the backbone of several important applications:

• Diffusion-weighted imaging (DWI): Measures water molecule diffusion in tissues. Restricted diffusion (bright on DWI) is a hallmark of acute stroke.
• Perfusion imaging: Measures blood flow using techniques like arterial spin labeling (ASL) or dynamic susceptibility contrast.
• fMRI: Maps brain activity by detecting blood oxygenation level-dependent (BOLD) contrast changes.

The tradeoff is that EPI is highly susceptible to artifacts from field inhomogeneities and chemical shift effects.

Parallel Imaging Techniques

Parallel imaging uses data from multiple receiver coils simultaneously to reconstruct images from undersampled k-space data. This reduces the number of phase-encoding steps needed, which directly speeds up acquisition.

Two widely used methods:

• SENSE (SENSitivity Encoding): Works in image space. It uses pre-measured coil sensitivity maps to "unfold" aliased images that result from undersampling.
• GRAPPA (GeneRalized Autocalibrating Partially Parallel Acquisitions): Works in k-space. It estimates missing k-space lines using calibration data and the spatial encoding inherent in the coil array.

The acceleration factor (typically 2–4x) reduces scan time, limits motion artifacts, and improves temporal resolution. This is especially valuable for cardiac MRI (where breath-holding time is limited) and abdominal imaging.

MRI Image Interpretation and Artifacts

Image Contrast and Weighting

The contrast you see in an MRI image depends on how the pulse sequence parameters (TR, TE, TI) weight the contributions of T1, T2, and proton density:

• T1-weighted images (short TR, short TE): Fat appears bright, water/CSF appears dark. These provide excellent anatomical detail and are the standard for post-contrast imaging (e.g., gadolinium-enhanced studies highlight areas of blood-brain barrier breakdown in tumors).
• T2-weighted images (long TR, long TE): Water/CSF appears bright, fat appears intermediate. These are highly sensitive to pathology since most disease processes increase tissue water content (edema, cysts, tumors).
• Proton density-weighted images (long TR, short TE): Contrast depends primarily on the concentration of hydrogen protons. Fluid appears bright, fat appears intermediate. These are particularly useful for evaluating joint structures (e.g., meniscal tears, rotator cuff injuries).

Common MRI Artifacts

Understanding artifacts helps you avoid misdiagnosis and improve image quality.

Motion artifacts result from patient movement or physiological motion (breathing, cardiac pulsation). They appear as blurring or ghosting along the phase-encoding direction.

Susceptibility artifacts arise from differences in magnetic susceptibility between adjacent tissues or from metallic objects. They cause signal loss or geometric distortion, and are most prominent at air-tissue interfaces (e.g., near the sinuses) and around metallic implants (dental fillings, surgical clips). These artifacts worsen at higher field strengths and with GRE sequences.

Chemical shift artifacts occur because fat and water protons resonate at slightly different frequencies (~3.5 ppm apart). This frequency difference causes spatial misregistration at fat-water boundaries, appearing as bright or dark bands at the edges of organs (e.g., around the kidneys, orbits, vertebral bodies). Increasing receiver bandwidth reduces this artifact but at the cost of SNR.

Aliasing (wrap-around) artifacts happen when anatomy outside the field of view (FOV) gets mapped into the image because of insufficient sampling in the phase-encoding direction. The tissue "wraps around" to the opposite side of the image.

Zipper artifacts appear as a bright or dark line running across the image, caused by RF interference from external sources or faulty RF shielding of the scanner room. Proper Faraday cage maintenance and eliminating electronic interference sources are the primary fixes.