How to Learn NMR and MRI: From Nuclear Spin to Molecular Structure and Medical Imaging

Wait, What? MRI Does Not Take a Photograph Inside the Body

An MRI scanner does not shine visible light into the body. It applies a strong static magnetic field, radiofrequency fields and magnetic-field gradients. The body then produces a weak radiofrequency signal, and computers reconstruct that signal into an image.

nuclear spin state → RF excitation → relaxation signal → spatial encoding → Fourier reconstruction → image contrast

The same tissue can look bright, dark or intermediate depending on the pulse sequence. MRI does not reveal one absolute appearance of tissue.

The One-Sentence Answer

Learn NMR and MRI by first understanding nuclear precession and relaxation, then use frequency to encode chemistry in NMR and magnetic-field gradients to encode position in MRI.

Stage 1: Only Some Nuclei Are NMR-Active

Nuclei with non-zero spin can possess magnetic moments. Common NMR nuclei include ¹H, ¹³C, ¹⁵N, ³¹P and ¹⁹F. MRI most commonly uses ¹H because hydrogen is abundant in water and fat.

Stage 2: A Static Magnetic Field Creates a Small Population Imbalance

Place many nuclear spins in B₀. Their allowed energy states are no longer equivalent. At thermal equilibrium, slightly more nuclei occupy the lower-energy state. Across enormous numbers of nuclei, that tiny imbalance creates measurable net magnetisation.

Stage 3: Spins Precess at the Larmor Frequency

A magnetic moment in B₀ precesses. The frequency obeys ω₀ = γB₀. The gyromagnetic ratio γ depends on the nucleus. A stronger field gives a higher resonance frequency.

Stage 4: Spin Is Not a Tiny Ball Literally Rotating

Nuclear spin is an intrinsic quantum property. The classical precessing-arrow model is useful for ensemble magnetisation, but should not be interpreted as a literal nucleus spinning like a planet.

Stage 5: Radiofrequency Excitation Tips Net Magnetisation

Apply an RF magnetic field near the Larmor frequency. Energy is transferred into the spin system and the net magnetisation can be rotated away from the B₀ axis. A 90-degree pulse is a controlled ensemble rotation under specified conditions.

Stage 6: The Rotating Frame Simplifies the Motion

In the laboratory frame, magnetisation precesses rapidly. In a frame rotating near the Larmor frequency, the RF interaction is easier to visualise. The rotating frame is a coordinate transformation.

Stage 7: After Excitation, the System Relaxes

Once RF excitation ends, nuclear magnetisation returns toward equilibrium. Two major processes are T1 longitudinal recovery and T2 transverse dephasing or decay.

Stage 8: T1 Is Longitudinal Recovery

T1 describes recovery of magnetisation along B₀. Energy is exchanged between the spin system and its surroundings. Different molecular environments produce different T1 values.

Stage 9: T2 Is Transverse Coherence Loss

After excitation, spins begin with phase coherence in the transverse plane. Interactions cause them to dephase. The collective transverse signal decays. T2 characterises this coherence loss.

Stage 10: T2* Includes Additional Field Inhomogeneity

Real magnetic fields are not perfectly uniform. Static variations cause extra dephasing. T2* is therefore generally shorter than T2. A spin echo can refocus some reversible inhomogeneity.

Stage 11: A Spin Echo Separates Reversible Dephasing From T2 Decay

A 180-degree RF pulse can reverse phase ordering caused by static field offsets so spins reconverge into an echo. True T2 interactions are not completely reversed.

Stage 12: NMR Spectroscopy Uses Frequency as Chemical Information

Electrons around a nucleus partially shield it from the external field. Different chemical environments therefore shift resonance frequency. This produces the chemical shift.

Stage 13: Chemical Shift Uses a Reference Scale

Chemical shifts are commonly reported in parts per million, making them more portable across field strengths. Solvent and reference conditions still matter.

Stage 14: Spin–Spin Coupling Adds Neighbour Information

Nuclei can influence one another through chemical bonds, producing multiplet splitting. A signal can reveal both its own environment and neighbouring spin topology.

Stage 15: Peak Area Can Be Quantitative

Under suitable acquisition conditions, integrated NMR peak area can be proportional to the number of contributing nuclei. Quantitative NMR can therefore measure amount or purity. NIST maintains SI-traceable qNMR reference materials.

Stage 16: Fourier Transformation Converts Time Signal Into Spectrum

After an RF pulse, the instrument records a decaying oscillatory free-induction signal containing many frequencies simultaneously. Fourier transformation converts it into a frequency-domain spectrum.

Stage 17: Higher Magnetic Field Can Improve Several NMR Properties

Higher B₀ can increase signal-to-noise and spectral dispersion, but can also amplify susceptibility effects, engineering cost and RF complexity. Higher field is useful, not universally superior.

Stage 18: Multidimensional NMR Spreads Information Across More Than One Axis

2D and higher-dimensional experiments correlate nuclei. They can reveal proton–proton coupling, proton–carbon connectivity and through-space proximity. A crowded 1D spectrum becomes a network of relationships.

Stage 19: Protein NMR Measures Structures as Dynamic Ensembles

Proteins in solution tumble and fluctuate. NMR can report chemical environment, distances, dynamics and conformational exchange. The output is not one frozen protein pose.

Stage 20: Solid-State NMR Solves a Different Motion Problem

In solids, orientation-dependent interactions are not averaged by rapid molecular tumbling. Methods such as magic-angle spinning rotate the sample to average selected anisotropic interactions.

Stage 21: MRI Adds Spatial Encoding to NMR

Ordinary NMR asks what frequencies exist. MRI adds where the signal came from. Magnetic-field gradients make resonance frequency or phase depend on position.

Stage 22: Slice Selection Uses Gradient Plus RF Bandwidth

Apply a field gradient so different positions have different Larmor frequencies. Transmit an RF pulse over a chosen frequency range and only spins within the corresponding slice are strongly excited.

Stage 23: Frequency Encoding Maps Position Into Frequency

During readout, a gradient makes position correspond to frequency. Fourier reconstruction converts the resulting spatial-frequency signal into image position.

Stage 24: Phase Encoding Maps Position Into Phase

Apply a gradient briefly so spins at different positions accumulate different phase shifts. Repeat the measurement with different phase-encoding strengths to reconstruct another image dimension.

Stage 25: K-Space Is Not a Blurry Version of the Image

MRI raw data are stored in k-space. Each point describes spatial-frequency content of the whole image. Central k-space strongly influences broad contrast and overall signal, while outer k-space strongly influences fine detail and edges.

Stage 26: Image Contrast Is Engineered With Timing

Sequence parameters such as TR, TE and TI change how strongly T1, T2 or other processes influence signal. This is why one anatomy can look radically different across sequences.

Stage 27: T1-Weighted and T2-Weighted Are Not Direct T1 and T2 Maps

A weighted image combines several effects. A quantitative T1 map estimates a physical parameter. A bright pixel in a T1-weighted image is not a direct numerical T1 measurement.

Stage 28: Diffusion MRI Adds Brownian-Motion Sensitivity

Magnetic-field gradients can sensitise signal to molecular displacement. Water diffusing freely loses phase coherence differently from water constrained by tissue microstructure.

Stage 29: Diffusion Tensor Imaging Is a Model of Directional Diffusion

In white matter, water diffusion can be anisotropic. DTI represents diffusion using a tensor. A 2026 review emphasises both its power and methodological limitations. A diffusion tensor is not a direct image of axons.

Stage 30: Fibre Tractography Is an Inference

Software integrates local diffusion-orientation estimates into streamlines. Those tracks can resemble neural pathways, but crossings, branching and limited resolution create ambiguity. Tractography produces plausible paths, not photographs of individual fibres.

Stage 31: Functional MRI Uses a Haemodynamic Proxy

The common BOLD method measures a blood-oxygenation-related signal. Neural activity changes blood flow, oxygenation and deoxyhaemoglobin concentration.

BOLD signal ≠ direct measurement of neuronal firing

Stage 32: Magnetic Susceptibility Creates Both Contrast and Artefact

Susceptibility differences near air, metal, blood products and mineral deposits alter phase and signal. Susceptibility-weighted imaging and quantitative susceptibility mapping use this deliberately; other sequences experience it as distortion.

Stage 33: MR Spectroscopy Brings NMR Chemistry Into Living Tissue

Magnetic resonance spectroscopy can resolve selected metabolites in a spatially defined tissue region, although overlapping peaks and low concentrations make quantification challenging.

Stage 34: Arterial Spin Labelling Uses Blood Water as an Endogenous Tracer

ASL magnetically labels inflowing blood water. No injected contrast agent is required. Later signal can estimate perfusion. Ordinary water becomes a tracer because its spin state was altered.

Stage 35: Hyperpolarisation Amplifies Weak Nuclear Signals

At thermal equilibrium, nuclear polarisation is tiny. Hyperpolarisation creates much larger population imbalances. For ¹³C-labelled molecules, this can enable real-time metabolic measurements. Quantification still requires careful RF and kinetic modelling.

Stage 36: Magnetic Resonance Fingerprinting Changes Acquisition Strategy

Magnetic resonance fingerprinting varies sequence parameters deliberately. The measured signal trajectory is compared with a model or learned representation to estimate multiple quantitative tissue properties. A 2025 Nature Protocols paper described deep molecular MRF for proteins, metabolites and pH-related effects.

Stage 37: MRI Does Not Use Ionising Radiation

MRI does not use X-rays or gamma rays. But non-ionising does not mean physically risk-free. The FDA highlights hazards from strong static magnetic fields, rapidly switching gradients, RF heating and incompatible implants or objects.

Stage 38: Gadolinium Contrast Is Separate From MRI’s Basic Signal

Many MRI scans require no contrast agent. When gadolinium-based agents are used, they alter local relaxation properties. The metal is not simply photographed.

Stage 39: Calibration Is Essential for Quantitative MRI

Quantitative MRI aims to measure physical parameters reproducibly across scanners, sites and time. NIST developed an MRI system phantom with SI-traceable reference properties to test relaxation accuracy, geometry, resolution and field-related distortions.

Stage 40: Professional MRI Is a Forward-and-Inverse Model Problem

Which spin dynamics generated this raw signal, which gradients encoded space, which reconstruction assumptions were applied, and which tissue property can legitimately be inferred from the final image?

Evidence: How Do We Know NMR Measures Molecular Environment?

Evidence comes from chemical-shift standards, coupling patterns, isotope substitution, known molecular structures and quantum calculations. In MRI, phantoms and controlled pulse sequences show that contrast changes predictably with T1, T2, diffusion and susceptibility.

Misconceptions Worth Hunting

MRI is a giant camera.
Nuclear spin means the nucleus literally spins like a ball.
T1 and T2 are the same recovery process.
MRI directly measures tissue colour.
K-space is a low-resolution image.
Diffusion tractography shows individual nerve fibres.
fMRI directly measures neurons firing.
MRI has no risks because it has no ionising radiation.

Transfer Check

A proton experiences a stronger B₀ field. What happens to Larmor frequency? It increases.

Two protons in different chemical environments experience different shielding. What appears in NMR? Different chemical shifts.

Add a magnetic gradient across a patient. What new information becomes available? Position can be encoded.

Finally, an fMRI region becomes brighter during a task. Can you conclude neuronal firing increased by a precisely proportional amount? No.

How We Know the Learning Has Held

A learner should be able to explain nuclear spin and magnetic moment; Larmor precession and RF excitation; T1, T2 and T2*; spin echoes; chemical shift and coupling; Fourier transformation; NMR spectroscopy versus MRI; gradients and spatial encoding; k-space; weighted images versus quantitative maps; diffusion MRI and BOLD limitations; and MRI safety and calibration.

Model Limits

The classical magnetisation vector is an ensemble model. Relaxation times depend on field strength and tissue environment. Simple mono-exponential relaxation can fail in complex tissues. DTI compresses complicated microstructure into one tensor. BOLD neurovascular coupling varies across state and region. Quantitative MRI depends on calibration and sequence modelling. Professional magnetic resonance keeps nucleus + field + pulse sequence + spatial encoding + reconstruction + calibration visible together.

Teaching Guide

Teach in this order: nuclear spin → B₀ → Larmor → RF pulse → T1/T2 → FID/Fourier → chemical shift → coupling → NMR structure → gradients → slice/frequency/phase encoding → k-space → MRI contrast → diffusion/fMRI → quantitative calibration.

Begin with: “If MRI is an image, where are the image pixels before reconstruction?”

At advanced level, compare an FID, NMR spectrum, k-space dataset and reconstructed MRI. Ask which transformation connects each representation to the next.

Connect This to the eduKate Learning Estate

Research Foundations and Further Learning

The Quiet Ending

The beginner asks, “Why do nuclei produce a signal in a magnet?” The developing physicist asks, “What frequency and relaxation process created this waveform?” The advanced learner asks, “How were chemistry or position encoded into the signal?”

Which spin-physics model and calibration chain makes the reconstructed spectrum or image a defensible measurement of molecular or tissue state?