← Radiology

Guide 8 min read

Metal, beam hardening, and the streaks across your scan

Why crowns and implants throw dark bands across a CBCT, what those artifacts actually are, and what to do about them before you over-call or under-call.

You have seen it. A patient with a couple of full-gold crowns gets a CBCT for something unrelated, and when you scroll to the molar region the bone has vanished into a fan of dark bands and bright spokes. The periapical area you actually wanted to see is sitting under a streak. The instinct is either to call the lucency you think you see or to ignore the region entirely. Both are traps, and understanding why the streak is there tells you which mistake you are about to make.

Beam hardening, in plain terms

An x-ray beam is not one energy. It is a spectrum of photon energies, and the low-energy photons are the easiest to absorb. As the beam passes through anything dense, those low-energy photons get stripped out first, so the beam that comes out the other side is, on average, higher energy. We call that a harder beam, and a harder beam is more penetrating. The problem is that the reconstruction math assumes the beam stays the same as it crosses the patient. It does not. So the machine mismeasures how much the dense material actually attenuated, and it paints that error into the image.

That error shows up two ways. The first is cupping, where a uniformly dense object reconstructs darker in its center than at its edges, because the central rays pass through more material and harden more. The second is the one that ruins molar reads: dark bands and bright streaks running between and around dense objects, worst along the line connecting two metal restorations. Even titanium, which is fairly low density as these things go, hardens the beam enough to streak. Gold, amalgam, and cobalt-chromium are worse.

It is not only beam hardening

Calling every metal streak “beam hardening” is a useful shorthand and a slightly wrong one. Schulze’s review of CBCT artifacts lays out the whole family, and the streaks you see are usually several of these at once.

Photon starvation is the big one near heavy metal. A gold crown two or three millimeters thick can absorb the great majority of the photons heading through it, on the order of ninety percent or more at the beam’s mean energy. When almost no signal reaches the detector along a given ray, the reconstruction has nothing reliable to work with, and that missing data reconstructs as severe bright and dark streaking. Push it to the extreme, where the signal is effectively zero, and you get what Schulze calls an extinction artifact, because the math behind reconstruction cannot take the logarithm of nothing.

Scatter is the other major contributor, and it is where CBCT really suffers. Scattered photons arrive at the detector having wandered off their true path, and they drag the measured attenuation down, cut contrast, and add their own dark streaks between dense objects. The scatter-to-primary ratio in cone beam imaging runs roughly forty times higher than in a collimated single-ray CT geometry. On top of those you get edge effects at sharp metal boundaries and partial volume averaging when a voxel straddles metal and bone. The practical point is that a metal streak is a pile-up of several physical problems, which is also why no single fix removes it cleanly.

Why CBCT gets hit harder than medical CT

If you have compared the same mouth on a hospital CT and a dental CBCT, the CBCT looks worse around metal, and there are concrete reasons for that. Dental CBCT runs at a lower tube current, roughly an order of magnitude below medical CT, so there are fewer photons to spare when metal starts eating them, and photon starvation arrives sooner. The tube voltage is often lower and fixed, commonly around ninety kilovolts, which leaves more soft, easily absorbed photons in the beam to begin with. The cone geometry and large flat panel capture far more scatter than the fan beam and tight collimation of a helical scanner. And many dental units lack the beam-shaping filtration, the bowtie filter, that medical scanners use to pre-harden and tame the beam. None of this is a defect in your machine. It is the trade you make for a low-dose, low-cost, chair-side volume.

What it does to the read

The danger is not that the image looks ugly. It is that the artifact imitates pathology in both directions. A dark band can read as a periapical lucency or a fracture line that is not there, and a bright streak can bury a real one. This is documented, not theoretical. In studies of vertical root fracture detection, teeth with metallic posts are exactly where readers and algorithms both stumble, because the artifact and the fracture live in the same space.

There is a specific, counterintuitive lesson buried in that literature. Metal artifact reduction software, the MAR setting on your console, does not always help. In one vertical root fracture study, turning MAR on lowered diagnostic accuracy compared with leaving it off, worst of all in the teeth with posts, and the authors suggested going back to a periapical radiograph as the next step rather than trusting the processed volume. MAR is an algorithm making its best guess at data the scan never truly captured, and near small dense objects that guess can erase a real finding or invent a smooth surface where there was a crack.

What to actually do

You cannot delete the physics, but you can keep it from fooling you.

Before the scan, remove what comes out. Removable partials and any loose metal in the field should be out of the mouth. Then match the field of view to the question and, when you can, position so the densest restorations are not sitting in the same plane as the region you care about. A smaller volume that excludes a mouthful of crowns is a cleaner volume. If your unit lets you raise the kilovoltage, a harder starting beam has fewer soft photons to lose to the metal, which reduces the beam-hardening component.

At the console, know what MAR is and is not. It genuinely helps for large high-density objects and can cut streak severity substantially, with one study reporting reductions well over half in standard mode. It also introduces its own artifacts, especially around small metal, and it can degrade specific tasks like fracture detection. Look at the scan both with MAR and without it when a metal-adjacent finding is on the line, and treat any lucency that appears in one but not the other with suspicion.

When you read, read the artifact as an artifact. Streaks that radiate from metal, follow the line between two restorations, and change as you scroll are telling you they are geometry, not disease. A true lucency has a location, a border, and a behavior across slices that an artifact does not. And when the metal genuinely buries the answer, the honest move is to say so and reach for a complementary view, a well-angled periapical or, for the right question, a different modality. A confident read of a region the physics has erased is worse than admitting the region was erased.

Bigger picture, this is the same lesson every part of imaging keeps teaching. The machine is not showing you the mouth. It is showing you a reconstruction of the mouth, built on assumptions, and where a mouthful of gold breaks those assumptions the picture lies in specific, learnable ways. Knowing the ways is most of the job.

Sources

Still have a question?

Ask me directly.

This page answers the version I get asked most. If your case is the exception, the inbox is faster than the website.

Get in touch →