Unlike X-ray, CT, or nuclear medicine, MRI uses no ionizing radiation. Instead, it leverages the magnetic properties of hydrogen protons in the body to create incredibly detailed images of soft tissues. This beginner's guide covers the MRI physics that every MR technologist must understand — from basic principles through practical sequence selection.
MRI doesn't produce images by measuring ionizing radiation transmitted through or emitted from the body. Instead, an RF magnetic field excites hydrogen nuclei, and receiver coils detect voltage induced by the changing transverse magnetization as the ensemble relaxes. Gradients spatially encode the signal. No ionizing radiation or rotating X-ray tube is involved—but the static, gradient, and RF fields have important safety hazards.
All of MRI rests on three physical phenomena. Understanding these is the foundation for everything else:
The main static magnetic field, measured in tesla (T). 1.5 T and 3 T are common clinically. Higher field can increase available SNR, but does not automatically improve resolution and may increase susceptibility, chemical-shift, RF-uniformity, and heating challenges.
An RF magnetic field near the Larmor frequency tips the net magnetization. A transmit-receive coil may do both jobs, but many examinations use the built-in body coil to transmit and a separate local array to receive.
Three sets of gradient coils create linear variations in the magnetic field along X, Y, and Z axes. They encode spatial position into the signal.
Hydrogen nuclei have spin angular momentum and a magnetic moment. In B₀, they occupy parallel and antiparallel energy states and precess about the field direction. At clinical temperature and field strength, only a tiny excess occupies the lower-energy state; the vector sum of the ensemble creates the measurable net magnetization along B₀. MRI descriptions such as “all protons align” are therefore useful shorthand, not a literal account of every nucleus.
Before excitation, individual transverse phases are distributed around the transverse plane and largely cancel. An on-resonance RF field (B₁) rotates the net magnetization by a flip angle determined by RF amplitude and duration. The changing transverse magnetization induces voltage in a receiver coil; MRI measures this ensemble signal rather than a photon emitted by each proton.
The equation at the heart of MRI is the Larmor equation:
f₀ = (γ/2π) × B₀
Angular frequency uses the distinct form ω₀ = γB₀, with ω₀ in radians per second and γ ≈ 267.5 Mrad/s/T for hydrogen.
At 1.5 T: Larmor frequency = 42.58 × 1.5 = ~63.9 MHz
At 3.0 T: Larmor frequency = 42.58 × 3.0 = ~127.7 MHz
Efficient RF excitation occurs near resonance. A pulse has a finite bandwidth; together with a slice-select gradient, that bandwidth determines which range of positions is excited. Different field strengths therefore use different center frequencies.
After RF excitation tips net magnetization away from equilibrium, longitudinal recovery and transverse dephasing occur simultaneously:
T1 is the time constant for longitudinal magnetization to recover 63% of the way from zero toward equilibrium after a 90° pulse. It involves energy exchange with the surrounding molecular environment.
T2 is the time constant for the transverse magnetization to decay to 37% of its initial value. It represents protons losing phase coherence due to spin-spin interactions.
T2* includes true T2 decay plus additional reversible dephasing from B₀ inhomogeneity and magnetic susceptibility, so T2* ≤ T2. A 180° refocusing pulse in a spin echo corrects much of the static dephasing and reveals T2 decay; gradient echoes do not provide that refocusing and are therefore T2*-sensitive. Neither echo can restore signal already lost through irreversible T2 processes.
| Material / finding | Typical T1-weighted | Typical T2-weighted | Typical FLAIR |
|---|---|---|---|
| CSF / Water | Dark ↓ | Bright ↑↑ | Dark ↓ (suppressed) |
| Fat | Bright ↑↑ | Intermediate | Intermediate |
| Gray matter | Gray | Gray | Gray |
| White matter | Light gray | Dark gray | Dark gray |
| Vasogenic edema | Usually low | Usually high | Usually high |
| Bone (cortex) | Dark ↓ | Dark ↓ | Dark ↓ |
| Blood products | Variable with age/state | Variable with age/state | Variable |
| Enhancing tissue after GBCA | Usually higher on T1-weighted imaging | Concentration/sequence dependent | Variable |
These are common appearances, not diagnostic absolutes. Signal depends on sequence, field strength, echo train, suppression pulses, tissue composition, flow, susceptibility, and contrast concentration. A tumor is not a single tissue type, and blood appearance evolves with hemoglobin state and time.
“T1 — fat is often bright; free water is usually dark.”
“T2 — free water is usually bright.” These aids apply to common non-fat-suppressed images; fat suppression, inversion recovery, gradient echo, and other sequence choices can reverse or null expected signal. Not every tumor or inflammatory focus is bright.
TR (Repetition Time) — Conventionally, the interval between repeated excitations of the same tissue. It strongly influences how much longitudinal recovery occurs, although its exact definition and contrast effect depend on the sequence.
The following conventional spin-echo ranges are approximate teaching examples, not universal cutoffs:
TE (Echo Time) — The time from excitation to the center of the measured echo. Longer TE permits more transverse decay; spin-echo TE emphasizes T2 effects, whereas gradient-echo TE emphasizes T2* effects.
| Sequence | TR | TE |
|---|---|---|
| T1-weighted | Short (400–800 ms) | Short (10–30 ms) |
| T2-weighted | Long (2000–5000 ms) | Long (80–120 ms) |
| PD-weighted | Long (2000–5000 ms) | Short (10–30 ms) |
| FLAIR | Long (6000–10000 ms) | Long (80–140 ms); long TI chosen to null CSF |
Without gradients, spins at different locations in a uniform B₀ would have the same Larmor frequency, so the received signal would lack spatial localization. The scanner has physical X, Y, and Z gradient sets, but it can combine them to create an oblique gradient. Slice-select, phase-encode, and readout are acquisition roles—not fixed coil identities.
Gradients are also responsible for the loud knocking sounds during MRI — the rapid switching of currents through gradient coils causes mechanical vibration (the Lorentz force on the coils).
RF coils may be transmit-only, receive-only, or transmit-receive. Many systems transmit with the integrated body coil and receive with a local multichannel array. Different designs are optimized for coverage, uniformity, acceleration, and SNR:
| Coil Type | Location | Best For |
|---|---|---|
| Body coil (built-in) | Inside the bore | Usually RF transmission; may also receive over a large FOV |
| Head coil | Encloses the head | Brain, orbits, sinuses; high local SNR |
| Spine coil | Embedded in table | Cervical, thoracic, lumbar spine |
| Surface / Phased-array | Placed on body surface | Knee, shoulder, breast, wrist — high resolution near the coil |
| Cardiac coil | Anterior + posterior array | Cardiac MRI, MR angiography |
Phased-array coils with multiple receiver elements allow parallel imaging (GRAPPA, SENSE) which reduces scan time by using multiple coil elements to partially replace gradient-encoded spatial information.
K-space is the sampled spatial-frequency data matrix, not a direct anatomical map. Its center is dominated by low spatial frequencies that strongly influence overall signal and contrast; its periphery contains higher spatial frequencies important for edges and detail. All samples contribute to every reconstructed pixel. A discrete Fourier transform converts the complex k-space data into image space.
Conventional 2D spin echo may acquire one phase-encoded line per TR, but this is not universal: fast/turbo spin echo acquires multiple lines in an echo train, and echo-planar imaging can traverse many or nearly all lines after one excitation. Partial Fourier, parallel imaging, compressed sensing, radial, and spiral methods alter how data are sampled and reconstructed.
There is no free improvement: changing a parameter usually affects more than one outcome. Smaller voxels improve nominal spatial resolution but reduce SNR. Thicker slices, larger in-plane pixels, and more averages can increase SNR, but may reduce resolution or increase time. SNR increases approximately with the square root of the number of signal averages, while acquisition time increases approximately linearly.
MRI uses no ionizing radiation, but “nonionizing” does not mean risk-free. The static field of a superconducting clinical scanner is normally always on; time-varying gradients and RF are present during acquisition. Follow the current ACR Manual on MR Safety, manufacturer instructions, device labeling, and facility policy.
The ACR four-zone model progresses from freely accessible Zone I, through the public-to-controlled interface and screening area in Zone II, to access-restricted Zone III, and the scanner room itself in Zone IV. Screen patients, accompanying people, and personnel before controlled access; repeat final checks according to site policy. Ferromagnetic detection supplements—but does not replace—a documented interview and device/foreign-body investigation.
For an implant or foreign body, record and verify the exact device, components, location, and conditions from reliable written/electronic sources. “The patient had MRI before” is not sufficient clearance. MR Conditional conditions may specify field strength, spatial field gradient, RF exposure (SAR or B₁+rms), gradient slew/dB/dt, coil configuration, scan duration, device programming, patient position, and monitoring.
The ARRT MR content specifications implemented February 1, 2025 list 200 scored questions: Patient Care 16, Safety 21, Image Production 106 (including 40 Physical Principles, 36 Sequence Parameters and Options, and 30 Data Acquisition/Processing/Storage), and Procedures 57. The examination also includes 30 unscored pilot questions. Use the current ARRT document rather than assuming an article's emphasis predicts exam percentages.
For more on how MRI compares to other modalities, see CT vs MRI: When to Use Which and explore our MRI modality overview.
Try these ARRT-style multiple choice questions based on this article. Click an option to check your answer — correct answers turn green, wrong ones turn red.