MRI Metal Artefact Reduction (MARS)

With the rising number of patients carrying orthopaedic implants, imaging around metal is an increasingly important skill. The core knowledge is which materials cause most artefact, why 1.5T is preferred, and what the basic MARS techniques do. Assessment of adverse local tissue reaction (ALTR) around metal-on-metal hips is a common application.

Where the artefact comes from. Metals have a different magnetic susceptibility from the surrounding tissue, so an implant distorts the local magnetic field. The metal itself returns no signal because it has no mobile protons, and the field distortion around it mismaps the signal from the tissue nearby.

What sets the severity. The implant's material, its alignment in the magnet, the spatial resolution and the sequence all change how much artefact an implant produces, and severity is proportional to field strength. Alignment is the one lever not covered in the sections below: placing the implant's long axis parallel to B0 reduces artefact.

View larger

Artefact follows the magnetic susceptibility of the material, from none with PEEK to severe with stainless steel.

Choosing the implant. When post-operative MRI is anticipated, choose titanium implants when possible. Cobalt-chrome is the difficulty in arthroplasty: metal-on-metal hip bearings cause significant artefact and need MARS sequences for ALTR surveillance, and total knee replacements with cobalt-chrome femoral components are also challenging to image.

Before any dedicated sequence is added, the ordinary parameters of the protocol can be set to limit artefact.

Why 1.5T. Metal artefact is directly proportional to field strength, and 3T produces approximately twice the artefact of 1.5T for the same implant. For routine imaging around metal, 1.5T is preferred despite its generally lower SNR. Modern multispectral MARS (MAVRIC-SL) can make 3T diagnostically useful (Choi 2015), but 1.5T remains the default where artefact is the limiting factor.

View larger

Why STIR, not chemical fat saturation. Chemical fat saturation relies on a uniform magnetic field to excite fat selectively, and the field inhomogeneity around metal makes it fail. STIR suppresses fat by inversion recovery, a T1 property, and so works regardless of field uniformity.

The problem is static. Near an implant the dominant problem is the large static off-resonance, the field inhomogeneity produced by the metal's susceptibility, and whether a sequence can undo the dephasing it causes decides how much artefact that sequence shows. General sequence physics is developed in the mri-imaging-principles topic; this section gives the metal-specific reason.

Why gradient echo fails. Gradient echo has no 180-degree refocusing pulse, so it cannot undo the phase that spins accumulate in the implant's static field inhomogeneity. The T2* sensitivity prized elsewhere for detecting haemosiderin or calcium becomes catastrophic blooming around metal.

Why spin echo works. The 180-degree pulse of spin echo and fast spin echo reverses the static dephasing and recovers the signal, and a short inter-echo spacing limits dephasing between refocusing pulses. Near any implant, image with FSE/TSE; SEMAC, MAVRIC and WARP are all built on spin-echo (VAT) backbones.

Distortion follows the frequency axes. Metal off-resonance mismaps signal along the frequency-encode (readout) direction in-plane and along the slice-select direction through-plane, because both axes use frequency to localise signal. The phase-encode direction is largely spared. That single fact explains the bandwidth rule, the value of swapping encoding directions, and why VAT and SEMAC are separate tools.

Receiver bandwidth. A higher receiver bandwidth spreads more hertz across each pixel, so a given off-resonance shift displaces signal by fewer pixels: less in-plane distortion and a smaller void. The price is signal-to-noise, which falls roughly with the square root of bandwidth, so bandwidth is increased when SNR allows.

View larger

Steering the artefact. Because the in-plane displacement runs along the frequency-encode axis, swapping the frequency- and phase-encode directions moves the pile-up and void off the structure you care about.

Why VAT and SEMAC are different tools. View-angle tilting corrects the in-plane, readout-direction distortion with an extra readout gradient. SEMAC adds slice-direction (z) phase encoding to correct the through-plane distortion. Combining them, with MAVRIC's multi-frequency acquisition, covers both directions, which is why these are the definitive techniques for large cobalt-chrome arthroplasty when the simple levers are not enough.

A mass is not automatically a pseudotumour. A solid or cystic periprosthetic mass must always be correlated with the bearing type, serum metal ions, inflammatory markers and joint aspiration. Infection and, rarely, sarcoma are the dangerous mimics, and they change management entirely.

View larger

View larger

View larger

View-angle tilting (VAT). VAT applies an additional gradient during readout that tilts the view angle, correcting in-plane geometric distortion caused by the metal. It is effective but increases scan time, and it is often combined with increased bandwidth.

SEMAC. Slice-encoding for metal artefact correction addresses the through-plane (slice-direction) distortion by acquiring multiple z-phase encoding steps for each slice, which resolves signal pile-up and slice distortion. The cost is a significant increase in scan time, although in knee arthroplasty compressed-sensing SEMAC has brought it under 5 minutes per sequence (Fritz 2016).

MAVRIC. Multi-acquisition variable-resonance image combination acquires images at multiple frequencies to cover the range of off-resonance the metal causes, then combines the sub-images into an artefact-reduced composite. It is effective for large metal implants, at the cost of long scan times. MAVRIC-SL combines it with SEMAC.

View larger

WARP. A Siemens proprietary technique that combines VAT with optimised bandwidth and slice-profile correction. It is faster than SEMAC or MAVRIC with good artefact reduction, and is available as a standard option on many Siemens scanners.

A smaller void is not the whole answer. MAVRIC reduces the component-adjacent void more than a conventional MARS set, but its images are more blurred, so choose the advanced sequence for the diagnostic target rather than for the smallest void.

View larger

View larger

Compressed-sensing SEMAC has its own false positives. It can show marrow that VAT obscures, but it can also create apparent muscle oedema distant from the implant, and its ripple artefact can mimic lamellated synovitis or loosening. Read corrected images with the conventional sequence: a finding that tracks the ripple pattern, or disappears on another plane or sequence, is artefact until proved otherwise.

View larger

View larger

When standard MRI is adequate. Small titanium screws and plates often produce acceptable images with standard optimised protocols (increased bandwidth, thin slices, spin echo). MARS sequences add scan time and are most valuable for large cobalt-chrome implants such as hip replacements, which is where full 3D SEMAC or MAVRIC is reserved.

View larger