An MRI (magnetic resonance imaging) scanner builds a detailed cross-section of the body using nuclear magnetic resonance — no X-rays involved. This interactive walkthrough builds the whole machine from scratch: a strong magnetic field B₀ lines up the hydrogen nuclei (protons) in your tissue, a radio pulse at the Larmor frequency f = γ·B₀ tips them, and the tiny signal they give off as they relax is turned into an image. You learn why the scanner needs a superconducting magnet, what T1 and T2 relaxation are, and how magnetic gradients plus a Fourier transform reconstruct the picture.
An MRI scanner uses a strong magnetic field to line up the hydrogen nuclei in your body, then sends in a radio pulse at the Larmor frequency to knock them out of alignment. As the nuclei relax back, they induce a small signal in a coil; gradient magnets tag each position with its own frequency, and a Fourier transform turns those signals into an image.
The Larmor frequency is the rate at which hydrogen nuclei precess (wobble) around the main field B₀, given by f = γ·B₀ with γ ≈ 42.58 MHz/T for hydrogen. At a clinical 1.5 T field that works out to about 63.9 MHz, which is why MRI scanners excite protons using radio waves.
T1 is the time for the magnetisation to recover along the main field B₀ (longitudinal relaxation), while T2 is the time for the spins to lose phase coherence in the transverse plane (transverse relaxation). Different tissues have different T1 and T2 values, and choosing the repeat time TR and echo time TE turns those differences into image contrast.
Because lining up the hydrogen nuclei needs a 1–3 tesla field, and a normal copper coil would melt from I²R heating long before reaching it. A niobium-titanium superconductor cooled by liquid helium to about −269 °C has zero resistance, so a huge current can circulate without wasting power as heat.
Gradient coils add a position-dependent field on top of B₀, so each location precesses at its own Larmor frequency and gets a unique frequency tag. The scanner records these signals as wave patterns in k-space, and a Fourier transform decomposes them into the grid of pixels that forms the final image.