Wait, What? MRI Does Not Take a Magnetic Photograph of the Body
An MRI scanner never points at a voxel and directly reads “this is brain” or “this is tumour”. It prepares nuclear spins, lets them evolve in magnetic fields, records radiofrequency signals and reconstructs a spatial image.
spin physics → RF excitation → relaxation and precession → spatial encoding → detected signal → reconstructed image
The image is a model-derived measurement of nuclear magnetic behaviour, not a direct photograph.
The One-Sentence Answer
Learn NMR and MRI by first understanding nuclear spin and Larmor precession, then use RF pulses and relaxation to explain signal formation before adding magnetic-field gradients to encode position and reconstruct images.
Stage 1: Selected Nuclei Behave Like Tiny Quantum Magnetic Moments
Nuclei with non-zero spin possess magnetic moments. Hydrogen-1 is especially important because it is abundant in water and fat and has favourable NMR sensitivity.
Stage 2: A Strong Magnetic Field Splits Spin Energy States
Place spins in a static field B₀. Nuclear magnetic moments interact with the field and the spin energy levels separate.
Stage 3: The Population Difference Is Tiny but Measurable
At thermal equilibrium, slightly more spins occupy the lower-energy state than the higher-energy state. That small imbalance produces a net magnetisation along B₀.
Stage 4: Larmor Frequency Links Field Strength to Precession
Spin magnetisation precesses at an angular frequency approximately ω₀ = γB₀. The gyromagnetic ratio γ depends on nuclear species.
Stage 5: Radiofrequency Excitation Resonates With the Spins
An RF field B₁ oscillating near the Larmor frequency can transfer energy and rotate the net magnetisation away from the longitudinal axis.
Stage 6: A 90° Pulse Is an Idealised Rotation
Pulse duration and amplitude determine flip angle. A nominal 90° pulse rotates equilibrium magnetisation into the transverse plane under ideal conditions.
Stage 7: The Measured NMR Signal Comes From Transverse Magnetisation
Precessing transverse magnetisation induces an oscillating voltage in the receiver coil. The time-domain signal is called the free induction decay.
Stage 8: Fourier Transform Converts Time Into Frequency
Different resonance frequencies interfere in the time-domain waveform. Fourier transformation separates them into a frequency spectrum.
FID in time → Fourier transform → spectrum in frequency
Stage 9: Chemical Shift Makes NMR a Chemistry Tool
Electrons shield nuclei from the applied field. Different chemical environments create slightly different effective fields and therefore different resonance frequencies.
Stage 10: Chemical Shift Is Reported in ppm
Parts per million normalises frequency differences by instrument field strength, allowing spectra from different magnets to be compared.
Stage 11: J-Coupling Reveals Bond Connectivity
Nuclear spins interact indirectly through bonding electrons. Scalar coupling can split resonances into multiplets and provide structural information about neighbouring nuclei.
Stage 12: Peak Area Can Report Relative Nuclei
Under appropriate quantitative conditions, integrated signal can relate to the number of contributing nuclei. Relaxation delays and pulse conditions matter.
Stage 13: T1 Describes Longitudinal Recovery
After excitation, longitudinal magnetisation recovers toward equilibrium with characteristic time T1. T1 depends on molecular motion and local magnetic interactions.
Stage 14: T2 Describes Transverse Dephasing
Transverse spins gradually lose phase coherence through spin–spin interactions. Signal decays with characteristic T2.
Stage 15: T2* Includes Field Inhomogeneity
Observed free-induction decay is often faster than intrinsic T2 because static field variations create additional dephasing. T2* combines true transverse relaxation and inhomogeneity.
Stage 16: Spin Echo Refocuses Static Dephasing
A 180° pulse can reverse phase dispersion caused by static field differences and create an echo. This separates reversible inhomogeneous dephasing from irreversible T2 processes.
Stage 17: NMR Spectroscopy and MRI Use the Same Core Spin Physics
NMR spectroscopy primarily resolves chemical frequency information. MRI adds controlled spatial encoding so signal can be assigned to locations.
Stage 18: Magnetic-Field Gradients Encode Position
Add a gradient and Larmor frequency changes with position. Different spatial locations can then accumulate different phases and frequencies.
Stage 19: Slice Selection Uses Frequency Selectivity
Apply a gradient while transmitting an RF pulse. Only spins whose local Larmor frequencies fall within the pulse bandwidth are excited strongly, selecting a slice.
Stage 20: Frequency and Phase Encoding Build Spatial Information
Additional gradients encode position along other directions. The scanner records samples in k-space rather than directly recording image pixels.
Stage 21: K-Space Is Spatial-Frequency Space
Central k-space strongly influences image contrast and broad structure, while outer k-space contributes fine spatial detail. A k-space point does not correspond directly to one image pixel.
Stage 22: Fourier Reconstruction Converts K-Space to Image Space
The measured spatial-frequency data are transformed mathematically to reconstruct the image.
measured encoded signal → k-space → inverse Fourier transform → image
Stage 23: MRI Contrast Is Sequence Dependent
T1-weighted, T2-weighted and proton-density-weighted images can show the same tissue differently. Image brightness is not a fixed inherent colour of tissue.
Stage 24: Repetition and Echo Times Shape Contrast
TR and TE determine how much T1 recovery and T2 decay influence the measured signal. Sequence timing becomes part of the tissue contrast mechanism.
Stage 25: Gradient Echoes Use Different Refocusing Physics
Gradient-echo sequences refocus using gradients rather than a 180° RF pulse and are sensitive to T2* effects. They are useful for rapid imaging and susceptibility-sensitive applications.
Stage 26: Diffusion MRI Measures Water Displacement Statistics
Diffusion-sensitive gradients make signal depend on molecular motion. Restricted and anisotropic diffusion can reveal tissue microstructure, particularly in white matter.
Stage 27: Diffusion Tensor Imaging Is a Model
DTI represents diffusion with a tensor and estimates preferred directions. Crossing fibres and complex tissue architecture can violate the simple tensor model.
Stage 28: Functional MRI Uses Blood-Oxygenation Contrast
BOLD fMRI depends on magnetic susceptibility differences associated with oxyhaemoglobin and deoxyhaemoglobin. Neural activity changes local metabolism and blood flow, producing an indirect haemodynamic signal.
BOLD ≠ direct measurement of neuronal firing
Stage 29: Magnetic Resonance Spectroscopy Measures Metabolite Resonances
MRS uses chemical-shift information within selected tissue regions to estimate metabolites such as choline, creatine and N-acetylaspartate.
Stage 30: Contrast Agents Change Relaxation
Gadolinium-based agents can shorten relaxation times near where they distribute, changing image contrast. Contrast enhancement reports altered relaxation and distribution, not direct molecular identity.
Stage 31: MRI Has No Ionising Radiation—but It Has Other Safety Physics
The static field attracts ferromagnetic objects, gradient switching can induce peripheral nerve stimulation and RF fields deposit energy as heat. Safety therefore involves screening, SAR limits and device compatibility.
Stage 32: Stronger Fields Bring Benefits and Challenges
Higher B₀ can improve signal and spectral separation, but also increase susceptibility effects, RF nonuniformity and technical complexity.
Stage 33: Parallel Imaging Uses Multiple Receiver Coils
Different coil sensitivity profiles provide extra spatial information, allowing undersampling of k-space and faster acquisition.
Stage 34: Compressed Sensing Uses Sparsity Assumptions
If the image or transformed representation is sufficiently sparse, fewer measurements can reconstruct an acceptable image under appropriate incoherent sampling and optimisation.
Stage 35: AI Reconstruction Adds Statistical Priors
Machine-learning reconstruction can reduce noise or accelerate imaging. It also adds assumptions learned from training data. Validation must check whether clinically important rare features are preserved.
Stage 36: Quantitative MRI Tries to Measure Physical Parameters
Instead of only weighted images, quantitative methods estimate T1, T2, diffusion coefficients, susceptibility or other parameters. The result still depends on model and calibration.
Stage 37: NMR Is Also a Structural Biology Tool
Solution NMR can determine protein structure, dynamics and molecular interactions without crystals. Chemical shifts, couplings and nuclear Overhauser effects provide constraints.
Stage 38: Relaxation Reveals Molecular Dynamics
NMR relaxation measurements can probe motions across picosecond-to-second timescales. The same phenomenon that causes signal loss becomes a molecular-dynamics measurement.
Stage 39: Professional Magnetic Resonance Is a Signal-Encoding Problem
Which spin population generated the signal, which pulse/gradient history encoded it, and which reconstruction and relaxation model converted that waveform into the reported spectrum or image?
Evidence: How Do We Know MRI Images Are Reconstructed From Spin Signals?
Phantom experiments with known geometry, gradient calibration, k-space manipulation and Fourier reconstruction show directly that changing spatial encoding changes image location and shape predictably.
Misconceptions Worth Hunting
- MRI directly photographs tissue.
- NMR and MRI use completely different physics.
- A 90° pulse permanently flips individual protons sideways.
- T1 and T2 are image brightness values.
- K-space is a blurry version of the image.
- BOLD fMRI directly measures neurons firing.
- No ionising radiation means no MRI safety concerns.
- AI reconstruction contains no extra assumptions.
Transfer Check
Increase B₀. What happens to Larmor frequency? It increases proportionally.
A spin-echo sequence refocuses field inhomogeneity. Does it remove intrinsic T2 decay? No.
A bright structure on a T2-weighted image appears dark on T1-weighting. Did the tissue physically change colour? No.
How We Know the Learning Has Held
A learner should be able to explain nuclear spin and magnetic moments; explain Larmor precession, RF excitation and FID; distinguish T1, T2 and T2*; explain spin echoes; explain chemical shift and J-coupling; explain gradients, slice selection, k-space and Fourier reconstruction; distinguish MRI weightings; explain diffusion MRI, BOLD and MRS; and identify major safety and model assumptions.
Model Limits
Simple vector models approximate quantum spin ensembles. Relaxation is often multiexponential in real tissue. DTI oversimplifies complex fibre geometry. BOLD depends on vascular physiology. AI reconstruction can introduce learned priors. Professional magnetic resonance keeps field strength + pulse sequence + relaxation + gradient encoding + reconstruction + biological interpretation visible.
Teaching Guide
Teach in this order: nuclear spin → B₀ → Larmor → RF pulse → FID → Fourier spectrum → T1/T2 → spin echo → gradient → k-space → image reconstruction → diffusion/BOLD/MRS → safety.
Begin with: “If MRI does not take a photograph, what physical signal is the scanner actually recording?”
Connect This to the eduKate Learning Estate
- How to Learn Magnetism and Electromagnetism
- How to Learn Medical Imaging
- How to Learn Spectroscopy
- How to Learn Quantum Measurement
Research Foundations and Further Learning
The Quiet Ending
The beginner asks, “Why does hydrogen make an MRI signal?” The developing physicist asks, “How did RF pulses and relaxation create the waveform?” The advanced learner asks, “How did gradients encode spatial position into k-space?”
Which spin physics, pulse sequence and reconstruction assumptions justify the tissue contrast or quantitative parameter reported in the final image?
Science Hub Route
Continue through the eduKate Sengkang Science Hub · Complete Science Index
