MRI Physics — Fundamentals and Principles

MRIninja Knowledge Base | Master / Reference Page Version 1.0 — July 2026

1. Introduction and Purpose of This Page

This page is the architectural hub for a dedicated MRI physics cluster on MRIninja. It is deliberately distinct from the existing MRI Parameters — Overview and Classification master page (9501), which documents the practical, operator-facing acquisition parameters (FOV, matrix, slice thickness, NEX, parallel imaging, and so on) — in other words, the knobs a technologist turns. This page instead documents the underlying physics that explains why those knobs behave the way they do: the nuclear phenomenon that makes MRI possible in the first place, how relaxation and spatial encoding actually work, how an image is mathematically reconstructed from raw data, and the safety physics that governs SAR and gradient limits.

This is a foundational, general-purpose reference page. It introduces core physical concepts at a survey level and lays out the full planned structure of the physics cluster (Section 4); it does not attempt to replace the dedicated child pages that will progressively cover each topic area in full technical depth.

2. Why MRI Physics Has Its Own Dedicated Cluster

Three clusters already exist on MRIninja that touch physics-adjacent territory, and each has a distinct, non-overlapping scope:

• MRI Sequences — Overview and Classification (9003): documents specific pulse sequences (TSE, GRE, EPI, STIR, DWI, and so on) — what each sequence is, when to use it, and its vendor names.
• MRI Parameters — Overview and Classification (9501): documents the acquisition parameters a technologist actively sets or adjusts, and their practical trade-offs (resolution vs SNR vs scan time).
• MRI Physics — Fundamentals and Principles (this page): documents the underlying physical phenomena — nuclear spin behaviour, relaxation mechanisms, spatial encoding theory, image reconstruction mathematics, hardware physics, and bioeffects/safety physics — that explain why the sequences and parameters in the other two clusters work as they do.

A technologist optimising a protocol needs the Parameters cluster; a physicist or a technologist wanting to understand why a given parameter change produces a given effect needs this cluster. The two are complementary and will cross-link extensively as physics child pages are built.

3. Core Physical Principles — Foundational Survey

3.1 Nuclear Spin, Precession, and the Larmor Equation

Atomic nuclei with an odd number of protons and/or neutrons — hydrogen-1 (a single proton) above all, given its natural abundance in water and fat — possess a quantum mechanical property called spin, which gives rise to a small magnetic moment. In the absence of an external magnetic field, these nuclear moments point in random directions and sum to zero net magnetization. When placed in a strong external magnetic field (B0), the phenomenon first independently observed by Felix Bloch and by Edward Purcell’s group in 1946 occurs: the nuclear moments precess around the direction of B0 at a frequency directly proportional to the field strength, given by the Larmor equation (ω₀ = γB0, where γ is the gyromagnetic ratio, a constant specific to each nuclear species) [1,2]. For hydrogen-1, this yields a precession frequency of approximately 42.58 MHz per tesla — the basis for every subsequent physical process in MRI, from RF excitation to signal reception.

3.2 Net Magnetization and RF Excitation

At thermal equilibrium in a strong field, a very slight statistical excess of nuclear moments aligns with B0 rather than against it, producing a small net longitudinal magnetization (Mz). This net magnetization is far too small to detect directly while aligned with B0; MRI signal detection instead relies on tipping this magnetization away from equilibrium using a radiofrequency (RF) pulse transmitted at the Larmor frequency — the resonance condition that gives the technique its name. The angle through which the magnetization is tipped (the flip angle) and the resulting precessing transverse magnetization are what actually induce a detectable signal in a receiver coil, as first demonstrated experimentally by Bloch, Hansen, and Packard, and by Purcell, Torrey, and Pound, in their landmark 1946 papers [1,2].

3.3 Relaxation: T1, T2, and T2*

Once tipped away from equilibrium, the magnetization does not remain there: it relaxes back toward its equilibrium state through two physically distinct, simultaneously occurring processes. T1 (spin-lattice) relaxation describes the recovery of longitudinal magnetization as energy is transferred from the excited spins back to the surrounding molecular lattice. T2 (spin-spin) relaxation describes the decay of transverse magnetization due to dephasing among spins caused by their mutual magnetic interactions. T2* relaxation is the transverse decay actually observed in a real scanner, and is always faster than true T2, because it also incorporates dephasing from macroscopic magnetic field inhomogeneities (both from imperfect magnet shimming and from local tissue susceptibility variation) in addition to the true molecular T2 process. Differences in T1 and T2 between tissue types are the fundamental source of essentially all soft-tissue contrast in MRI, and are themselves dependent on field strength, molecular environment, and (as documented on the Contrast Media in MRI master page) the presence of paramagnetic or superparamagnetic contrast agents.

3.4 Spatial Encoding and k-Space

A precessing nuclear magnetic moment produces a signal at a single frequency, with no inherent spatial information. Spatial localisation in MRI is achieved entirely through the controlled application of magnetic field gradients — small, linear, deliberately superimposed variations in field strength across the three spatial axes — which make the local Larmor frequency (and, for phase encoding, the local phase) a linear function of position. Raw MRI data is acquired not directly as an image but as a set of spatial-frequency samples in a mathematical space called k-space; the elegant “spin-warp” method of filling k-space using a phase-encoding gradient combined with a frequency-encoding (readout) gradient, described by Edelstein and colleagues in 1980, remains the conceptual basis of the great majority of clinical pulse sequences in use today [5]. Multi-planar, gradient-based slice-selective imaging — the ability to directly acquire any oblique plane without physically moving the patient, a defining practical advantage of MRI over other cross-sectional modalities — was described by Peter Mansfield in 1977 [4].

3.5 Image Formation and the Fourier Transform

Because k-space is mathematically the two- (or three-) dimensional spatial-frequency representation of the image, the final image is obtained by applying an inverse Fourier transform to the acquired k-space data. This mathematical relationship — first proposed as a practical basis for NMR imaging by Paul Lauterbur in his seminal 1973 Nature paper [3] — underlies essentially every image-formation and image-acceleration technique used in modern MRI, including parallel imaging techniques such as SENSE and GRAPPA, which exploit the spatial sensitivity information of multi-channel receiver coil arrays to reconstruct full images from deliberately undersampled k-space data [8,9].

4. Structure of the MRI Physics Cluster — Planned Topics

This section is the roadmap for the cluster. Each numbered group below is intended to become one or more dedicated child pages of this master; none of them are covered in full technical depth here.

4.1 Fundamentals of Nuclear Magnetic Resonance

• Nuclear spin, magnetic moment, and precession (the Larmor equation)
• The B0 field and net magnetization
• RF excitation and the rotating reference frame
• Free induction decay (FID)

4.2 Relaxation Phenomena

• T1 relaxation (spin-lattice)
• T2 relaxation (spin-spin)
• T2* and the role of field inhomogeneity
• Tissue- and field-strength-dependence of T1/T2

4.3 Signal Localisation and Image Formation

• Gradients: slice-select, phase-encode, frequency-encode/readout
• Physics of slice selection
• k-Space: theory, filling strategies, symmetry (partial Fourier), trajectories (Cartesian, radial, spiral)
• The Fourier transform and image reconstruction
• The resolution / SNR / scan-time relationship — physical basis

4.4 RF Pulses and Pulse Sequence Physics

• Flip angle and the Ernst angle
• Selective, non-selective, and adiabatic pulses
• Refocusing pulses and echo formation (spin-echo vs gradient-echo physics)
• Magnetization transfer

4.5 Chemical Shift and Fat/Water Physics

• The chemical shift phenomenon and the ppm scale
• Chemical shift artefact (type 1 and type 2)
• Physical basis of fat suppression (STIR vs spectral vs Dixon methods)

4.6 Flow, Motion, and Diffusion Physics

• Flow-related phenomena (time-of-flight and phase-contrast physics)
• Physical origin of flow artefacts
• Diffusion physics (Brownian motion, the b-value, ADC)
• Physiological gating and triggering physics

4.7 Magnetic Field, Hardware, and Homogeneity

• Magnet types (superconducting, permanent, resistive) and their physics
• Field homogeneity and shimming (passive and active)
• Gradient coils: slew rate, maximum amplitude, and performance trade-offs
• RF coils: transmit/receive physics, coil types, and the hardware basis of parallel imaging
• Multi-transmit and B1 shimming at high field

4.8 Physics of MRI Artefacts

• Susceptibility artefact
• Aliasing / wrap-around physics
• Truncation (Gibbs) ringing
• RF-related artefacts (e.g. zipper artefact)
• The magic angle effect

4.9 SAR, Bioeffects, and MRI Safety Physics

• RF power deposition and the physics of SAR
• Gradient dB/dt and peripheral nerve stimulation
• Static field bioeffects
• Acoustic noise physics

4.10 High-Field Physics

• Field-strength dependence of SNR
• T1/T2 changes with field strength
• Susceptibility effects at high field
• B1 inhomogeneity at high field

4.11 Reconstruction and Post-Processing Physics

• Mathematical/physical basis of SENSE and GRAPPA
• Compressed sensing
• Physical principles behind AI/deep-learning reconstruction

5. Relationship to Sequences, Parameters, and Protocol Pages

Every anatomical protocol page on MRIninja makes physics-dependent decisions implicitly — choosing STIR over spectral fat saturation off-isocentre (chemical shift and field-homogeneity physics), accepting a specific parallel imaging factor (hardware and reconstruction physics), or selecting a b-value for DWI (diffusion physics). As this cluster is built out, physics child pages will be cross-linked from the relevant sequence pages (9003 cluster), parameter pages (9501 cluster), and anatomical protocol pages wherever a genuinely strong, non-forced semantic connection exists — consistent with the site-wide rule against weak or forced linking.

6. A Brief History of MRI Physics

7. Evidence Gaps and Ongoing Debate

• Physical validity limits of aggressive acceleration: as compressed sensing and AI-based reconstruction push undersampling factors higher, the physical and statistical assumptions underlying faithful image reconstruction (sparsity, learned priors) are an active area of technical debate, particularly regarding the risk of reconstructing plausible-looking but physically unsupported detail (“hallucination”) at high acceleration factors — to be addressed in the planned Reconstruction and Post-Processing Physics child pages (Section 4.11).
• Ultra-high-field physics challenges: B1 inhomogeneity, increased susceptibility artefact, and SAR management become substantially more physically challenging above 3T, and consensus technical solutions (multi-transmit shimming strategies in particular) continue to evolve — to be addressed in the planned High-Field Physics child pages (Section 4.10).
• Standardisation of physics terminology across vendors: vendor-specific implementation differences (addressed in the Parameters cluster) frequently stem from genuinely different underlying physics/engineering choices (e.g. gradient coil design, RF pulse design) rather than pure naming variation; the planned Hardware Physics child pages (Section 4.7) will need to draw this distinction carefully and will flag where authoritative comparative technical data is limited.

9. Evidence-Based References

D. Technical MRI Papers