How MRI Machines See Inside You (Without Cutting You Open)

The Machine That Shouldn’t Exist

If I described an MRI machine to you without telling you what it was, you’d think I was making it up. Here’s a device that creates a magnetic field 60,000 times stronger than Earth’s, uses it to align billions of hydrogen atoms in your body along one direction, then flips them with a carefully tuned radio pulse and listens for the tiny radio signal they emit as they relax back to alignment. From that signal — measured at different positions, angles, and timings — it reconstructs a detailed three-dimensional map of your internal organs, distinguishing between grey matter and white matter in your brain, identifying tumours smaller than a centimetre, and doing all of this without making a single incision, without any ionising radiation, and without touching you.

It sounds absurd. But it works, and the physics behind it is a remarkable chain of ideas connecting quantum spin, nuclear magnetic resonance, Fourier transforms, and superconducting magnets.

The Starting Point: Hydrogen and Spin

Your body is roughly 60% water by mass. Water is H₂O — two hydrogen atoms for every oxygen. Hydrogen is the simplest atom: one proton, one electron. The proton has a quantum property called spin — intrinsic angular momentum — that gives it a tiny magnetic moment. Each hydrogen proton is, in effect, a microscopic bar magnet.

Normally, these tiny magnets are randomly oriented. Their magnetic moments point in all directions and cancel out. You’re not magnetic (your fridge doesn’t stick to you). But put the body in a strong external magnetic field, and something changes.

In a strong field, the proton spins can align either parallel to the field (low energy) or anti-parallel (high energy). Quantum mechanics says these are the only two options — no intermediate orientations. Slightly more protons align parallel than anti-parallel, because the parallel state has lower energy. The excess is tiny: in a 1.5 T field at body temperature, only about 5 extra protons per million align with the field compared to against it.

Five per million sounds pathetic. But your body contains about 10²⁸ hydrogen atoms. Five per million of 10²⁸ is about 5 × 10²² — fifty billion trillion aligned protons. That’s enough to produce a detectable signal. Not a strong one — MRI signals are faint, which is why the technology requires sensitive receivers and extensive signal averaging — but detectable.

The RF Pulse: Tipping the Alignment

Those 5 × 10²² aligned protons create a net magnetisation vector — a tiny magnetic field pointing along the direction of the scanner’s main magnet (conventionally called the z-axis). This magnetisation is static and doesn’t produce a detectable signal. To get a signal, you need to disturb it.

The scanner sends a brief pulse of radio-frequency electromagnetic radiation at a very specific frequency — the Larmor frequency — which depends on the magnetic field strength and the type of nucleus. For hydrogen protons in a 1.5 T field, the Larmor frequency is 63.87 MHz — in the FM radio band.

When the RF pulse is tuned to exactly the Larmor frequency, the protons absorb energy and their net magnetisation tips away from the z-axis. A 90° pulse tips the magnetisation into the transverse plane (perpendicular to the main field). A 180° pulse flips it completely upside down.

This is resonance — the same physics that makes a wine glass vibrate when you sing at its natural frequency, or a child on a swing go higher when you push at the right rhythm. You’re driving a system at its natural frequency, and it absorbs energy efficiently. The “nuclear magnetic resonance” in NMR (the spectroscopy technique that MRI descended from) refers to exactly this.

The Signal: Listening to Protons Relax

After the RF pulse ends, the protons are in an excited state — their net magnetisation is tipped away from the equilibrium direction. They want to return to alignment. This relaxation is not instantaneous; it takes tens to hundreds of milliseconds, depending on the tissue type.

As the magnetisation precesses (rotates around the z-axis at the Larmor frequency) and relaxes back to equilibrium, it acts like a tiny rotating magnet. A rotating magnetic field induces a voltage in a nearby coil — this is the MRI signal. The receiver coil picks up a faint, decaying oscillation — a “free induction decay” — that encodes information about the hydrogen protons in that region.

Here’s where it gets medically useful. Different tissues relax at different rates. Fat protons relax quickly (short T1 time). Water protons in cerebrospinal fluid relax slowly (long T1). Grey matter and white matter in the brain have different T1 values. Tumours typically have different relaxation times than healthy tissue.

By choosing specific timing parameters for the RF pulses and signal acquisition — the “pulse sequence” — the scanner can make images where the contrast depends on T1 relaxation, T2 relaxation, proton density, or various other physical properties. Each pulse sequence produces a different type of contrast, highlighting different tissue characteristics. This is why radiologists often acquire multiple sequences for each body part — each one shows different things.

Spatial Encoding: Where Is the Signal Coming From?

The RF signal from the whole body would be useless as an image — it would just be one averaged signal from everything. To create an image, the scanner needs to know where in the body each part of the signal is coming from. This is the hardest part of MRI physics, and the cleverest.

The scanner uses gradient coils — additional electromagnets that add a small, precisely controlled variation to the main magnetic field. If you make the field slightly stronger on the left side of the body and slightly weaker on the right, the Larmor frequency changes across the body. Protons on the left precess faster; protons on the right precess slower.

Now the signal from the left side of the body has a slightly different frequency than the signal from the right side. A Fourier transform separates these frequencies, revealing how much signal comes from each position. One gradient gives you one spatial dimension.

Three gradients — along three perpendicular axes — give you full three-dimensional spatial encoding. The slice-select gradient excites only a thin slab of tissue. The frequency-encode gradient labels positions along one axis by frequency. The phase-encode gradient labels positions along the other axis by the phase of the signal (acquired over multiple repetitions with different gradient strengths).

The raw data sits in what physicists call k-space — a two-dimensional Fourier domain where each point represents a spatial frequency component of the image. Fill k-space completely (which requires many RF pulses with different phase-encode steps), apply a 2D inverse Fourier transform, and out pops the image.

It’s computationally elegant and physically brutal — acquiring enough data to fill k-space takes minutes, during which the patient must hold still while the gradient coils bang and rattle through hundreds of switching cycles per second.

The Superconducting Magnet

The main magnet of an MRI scanner is a superconducting solenoid — a coil of niobium-titanium wire cooled to 4.2 K (−269 °C) with liquid helium. At this temperature, the wire has zero electrical resistance. Once current is established in the coil, it flows forever — literally. Some MRI magnets have been running on the same current for 20 years without being topped up.

The field must be extraordinarily uniform — variations of less than a few parts per million across the imaging volume. Achieving this requires careful coil design, manufacturing precision, and a process called shimming — placing small iron pieces and running correction currents through additional coils to fine-tune the field homogeneity.

The liquid helium is expensive and gradually boils off (though modern scanners have cold-heads that recondense most of it). A quench — the sudden loss of superconductivity — causes the stored magnetic energy to convert to heat instantly, boiling all the helium at once and filling the room with freezing gas. Quenches are dramatic, dangerous, and rare. Emergency quench buttons exist in every MRI suite for situations where the magnet must be shut down immediately (for example, if a person is pinned to the scanner by a ferromagnetic object).

The Projectile Problem

Speaking of ferromagnetic objects — this is the most viscerally dramatic part of MRI physics. A 3 T magnet exerts a force on ferromagnetic materials (iron, nickel, cobalt, some steels) that increases steeply as the object approaches the bore. A steel oxygen cylinder in the same room as an MRI scanner can be pulled toward the magnet with a force of hundreds or thousands of newtons — enough to fly across the room like a missile.

MRI safety incidents involving projectile objects are rare but headline-making. There have been cases of floor buffers, IV poles, and oxygen tanks being pulled into scanners. The scanner itself is never damaged — the magnet is much tougher than whatever hits it — but anyone between the projectile and the magnet is in serious danger.

This is why MRI suites have strict screening procedures: no metal objects past the controlled zone. Patients are screened for implants, fragments, and loose metal. Staff wear no watches, belts, or hairpins. It’s not paranoia. It’s 60,000 times Earth’s magnetic field, concentrated in a space the size of a person, running 24 hours a day whether there’s a patient inside or not.

From Physics Experiment to Medical Miracle

Nuclear magnetic resonance was discovered by Isidor Rabi in the late 1930s (Nobel Prize 1944). Felix Bloch and Edward Purcell independently developed NMR spectroscopy in 1945 (Nobel Prize 1952). Raymond Damadian demonstrated that NMR relaxation times differed between normal and cancerous tissue in 1971. Paul Lauterbur and Peter Mansfield developed the spatial encoding methods that turned NMR into imaging in the 1970s (Nobel Prize 2003).

The journey from “protons precess in a magnetic field” to “we can see a 3 mm tumour in your brain” took about 40 years. Every step required understanding the physics more deeply and engineering it more precisely. The modern MRI scanner is, arguably, the most physics-intensive device in routine use anywhere in the world — bringing together quantum mechanics, electromagnetism, superconductivity, signal processing, and an extraordinary amount of engineering into a machine that saves lives daily by seeing what no surgeon’s eye can see.