How Do MRI Machines Work?

An MRI machine maps the inside of your body using physics you learned in high school but probably never expected to see applied this way. There’s no radiation, no cutting, no contrast dye in most cases. Just a powerful magnet, carefully timed radio pulses, and hydrogen atoms doing exactly what quantum mechanics says they should.

The result is among the most detailed views medicine has of soft tissue: brain tumors smaller than a pea, torn knee ligaments, the precise boundary of a herniated disc. Understanding how it works isn’t just interesting physics. It explains why the machine is so large, why it costs millions of dollars, why you have to remove your keys, and why the noise is absolutely relentless.

The short answer

Your body is mostly water, and water is mostly hydrogen. MRI takes advantage of the fact that hydrogen atoms behave like tiny magnets. Inside the MRI’s powerful magnetic field, those hydrogen atoms align. A brief radio pulse knocks them out of alignment. As they snap back, they emit a faint radio signal. By varying the magnetic field across space, the machine can pinpoint exactly where each signal came from and reconstruct a detailed cross-sectional image.

The full picture

The magnet and what it does to hydrogen

The defining component of an MRI machine is its main magnet, a superconducting coil cooled to near absolute zero with liquid helium. Clinical MRI magnets run between 1.5 and 3 Tesla. Research machines go higher. To put that in perspective: Earth’s magnetic field is about 0.00005 Tesla. The MRI is roughly 60,000 times stronger.

This field is always on. The machine doesn’t switch it off between patients. Quenching a superconducting magnet (forcing it to suddenly lose superconductivity) releases all that energy at once, boiling the helium and potentially damaging the coil. Hospitals do it only in emergencies.

Inside that field, something specific happens to hydrogen nuclei. A hydrogen nucleus is a single proton, and protons have a property called spin: they behave as if they’re rotating, which gives them a tiny magnetic moment. Normally these spins point in random directions and cancel out. Inside a strong magnetic field, they preferentially align with the field, like compass needles pointing north. Not all of them align perfectly, and quantum mechanics dictates they don’t point exactly along the field. They precess around it, like a spinning top wobbling around the direction of gravity.

The rate of this precession is the Larmor frequency, and it depends on the magnetic field strength. At 1.5 Tesla, hydrogen protons precess at about 64 megahertz, which is in the radio frequency range. That’s what makes the whole technique work.

The radio pulse and resonance

To generate a signal, the machine broadcasts a radio frequency pulse at exactly the Larmor frequency. This is the “resonance” in magnetic resonance imaging. When you transmit at the exact frequency that hydrogen atoms are precessing, energy transfers efficiently. The protons absorb it and get knocked out of their alignment with the main field.

The moment you turn off the radio pulse, the protons start returning to their equilibrium alignment. As they do, they release the absorbed energy as a radio signal at the same frequency. This emitted signal is what the machine detects.

Here’s the critical detail: different tissues return to equilibrium at different rates. Fat protons relax quickly. Water protons in cerebral spinal fluid relax slowly. Muscle falls somewhere in between. By varying the timing of the radio pulses and the measurement windows, radiologists can generate images where different tissue types appear bright or dark. A T1-weighted image makes fat bright and fluid dark, ideal for anatomy. A T2-weighted image reverses this, making fluid bright, which is useful for spotting inflammation or tumors. These timing parameters are called pulse sequences, and radiologists choose among dozens of them depending on what they’re looking for.

Gradient coils and spatial encoding

The main magnet tells you that hydrogen is resonating somewhere in the patient’s body. To build an image, you need to know where. This is solved by gradient coils: electromagnets inside the bore of the machine that add a small, controlled variation to the magnetic field across the body’s three axes.

By varying the field slightly from head to toe, left to right, and front to back, you change the Larmor frequency at each location. Protons in a higher-field region precess slightly faster; protons in a lower-field region precess slightly slower. When you apply a gradient and then measure the returning signal, the frequency of the signal encodes the position along that axis.

Combining gradients across all three dimensions, and applying them in rapid sequence, gives you spatial information from every point in the imaging volume. A mathematical operation called a Fourier transform then converts this frequency-encoded data into the spatial image you recognize as a scan.

The gradient coils switching on and off many times per second is why MRI machines are so loud. The electromagnetic forces physically flex the coils against their mounting, and at clinical field strengths this creates loud banging, knocking, and rhythmic clunking that can exceed 100 decibels. Ear protection is not optional.

What contrast agents do

About a third of MRI scans use a contrast agent, most commonly gadolinium, a rare earth metal that’s injected intravenously. Gadolinium is paramagnetic, meaning it alters the local magnetic environment around it and causes nearby protons to relax faster. Tissues that absorb gadolinium appear brighter on T1-weighted images.

This is particularly useful for tumors and inflammation, which tend to accumulate contrast agents more than healthy tissue due to abnormal blood vessel permeability. Gadolinium doesn’t reveal tumors because it goes into the tumor. It reveals them because it preferentially concentrates there in ways healthy brain tissue doesn’t.

Gadolinium is generally safe, but researchers have found that trace amounts can accumulate in the brain and other tissues with repeated exposure. For patients with kidney disease, gadolinium carries additional risks. This is why radiologists carefully consider whether contrast is actually necessary for each scan.

The signal, the noise, and why stronger magnets cost more

Every MRI image is essentially a ratio of signal to noise. The radio signal from precessing hydrogen atoms is genuinely faint. You’re detecting the emissions from trillions of hydrogen nuclei, but the signal from any single one is vanishingly small. The main magnet’s job is partly to increase the fraction of protons that align with the field, which increases the available signal.

Stronger magnets mean more signal relative to noise, which means either faster scan times, higher resolution images, or both. A 3 Tesla scanner can see features that a 1.5 Tesla scanner might miss. Research scanners at 7 Tesla produce images with sub-millimeter resolution. The price difference is proportional: a 3T MRI costs roughly $3 million, compared to $1.5 million for a 1.5T machine.

The superconducting coils require liquid helium cooling to maintain temperatures near absolute zero. Recent developments in “zero boiloff” technology have reduced helium consumption significantly, but the infrastructure requirements (shielded rooms to contain the magnetic field, quench pipes to safely vent emergency helium release, rigid exclusion zones for anyone with ferromagnetic implants) explain why installing an MRI suite is as expensive as purchasing the machine itself.

Why it matters

MRI replaced a generation of exploratory surgeries. Before widespread MRI availability, diagnosing a meniscus tear meant an orthoscopic procedure to look inside the joint. A suspected brain tumor required a CT scan followed by a biopsy. MRI made a large category of soft tissue diagnoses non-invasive and non-irradiating, which changed clinical practice fundamentally.

The lack of ionizing radiation makes MRI the preferred modality for pediatric patients, pregnancy, and anyone requiring frequent surveillance scans. Brain imaging, spinal cord assessment, joint injuries, cardiac function, liver lesions: these are areas where MRI’s ability to discriminate between soft tissues gives it an advantage no other imaging modality can match.

The technology continues to advance. Ultra-high-field scanners at 7 Tesla are now approved for clinical use. Hyperpolarized gas MRI can image the airspaces in the lungs directly. Diffusion tensor imaging tracks water movement along nerve fibers to map white matter tracts in the brain. The same physical principle (hydrogen atoms in a magnetic field) keeps finding new clinical applications because the underlying physics is so rich with information.

Key terms

Larmor frequency The rate at which a proton precesses in a magnetic field, proportional to field strength. For hydrogen at 1.5 Tesla, this is approximately 64 MHz, the radio frequency that MRI exploits for resonance.

Precession The wobbling motion of a spinning object around an axis. Hydrogen protons precess around the direction of the main magnetic field, like a gyroscope wobbling around vertical.

T1 and T2 relaxation Two time constants describing how quickly protons return to equilibrium after a radio pulse. T1 is the time to realign with the main field; T2 is the time for protons to lose phase coherence with each other. Different tissues have different T1 and T2 values, enabling tissue contrast.

Gradient coils Electromagnets that impose a small, spatially varying field on top of the main magnet. By changing Larmor frequencies across space, they encode positional information into the detected signal.

Pulse sequence The specific pattern of radio pulses and gradient applications used to generate an image. Radiologists choose pulse sequences to highlight specific tissue properties, such as T1-weighted, T2-weighted, or diffusion-weighted imaging.

Fourier transform A mathematical operation that converts a signal from the time/frequency domain into the spatial domain. MRI uses it to translate frequency-encoded detector signals into spatial images.

Tesla The unit of magnetic field strength. Earth’s field is about 0.00005 Tesla. Clinical MRI machines run at 1.5 to 3 Tesla. The strong field is necessary to produce sufficient signal from the small fraction of protons that align with it.

Common misconceptions

“MRI and CT scans are interchangeable.” They’re not. CT uses X-ray radiation and excels at imaging bone and dense structures quickly. MRI uses radio waves and excels at soft tissue. A radiologist chooses based on what they’re looking for and the patient’s situation.

“The magnetic field turns off between uses.” It doesn’t. A superconducting MRI magnet runs continuously. The magnetic field extends several meters beyond the bore, which is why the room has strict entry protocols.

“Louder MRI means something is wrong.” The noise is intrinsic to gradient coil switching and varies by pulse sequence. Fast scanning sequences are louder. It’s expected, not a fault.

“Metal fillings prevent MRI.” Most modern dental fillings and crowns are titanium or ceramic, which aren’t ferromagnetic. The concern is implanted devices like pacemakers, cochlear implants, and certain aneurysm clips, not standard fillings. The radiologist screens for these specifically.