← Radiology

Guide 9 min read

Why your CBCT is not a medical CT

They share three letters and almost nothing else. Why cone beam is brilliant for bone, poor for a soft-tissue mass, and misleading when it reports a density number.

A referring dentist once sent me a CBCT with a note asking for the Hounsfield units at a planned implant site, because a lecture had told him a certain number meant good bone. I understood the request. It is a completely reasonable thing to want. It is also asking the machine for something it cannot honestly give, and the reason why runs straight through what a cone beam scanner actually is. It shares three letters with the CT scanner down the hall at the hospital, and past those three letters they are different instruments solving different problems.

Two different ways to photograph a head

A medical CT uses a fan-shaped beam and a row of detectors, and it builds the volume one slice at a time. The tube spirals around the patient while the table slides through, each rotation reconstructing a thin axial slice, and the slices stack into a three-dimensional whole. It is a careful, sequential process, and it is why a hospital CT is a big machine with a moving table.

A cone beam scanner does something almost cheeky by comparison. It sends out a single cone-shaped beam onto a flat panel detector, the same kind of area detector lineage as a digital camera sensor, and it captures the entire field of view in one rotation around the head. Scarfe and Farman, in the paper most of us learned this from, describe it plainly: the exposure incorporates the whole field, so only one rotational sequence is needed. During that single sweep the machine grabs somewhere between about a hundred fifty and six hundred flat projection images, the basis images, and from that stack it reconstructs the volume. One trip around the head, often in the ten to thirty seconds of a panoramic exposure, and you have a 3D dataset.

That architectural choice, one cone and one rotation instead of a fan and a spiral, is the root of nearly every strength and every weakness that follows.

What the single rotation buys, and what it costs

The reconstruction math is different too. Cone beam volumes are almost universally built with an algorithm called FDK, after Feldkamp, Davis, and Kress, which is fast and practical but approximate. It reconstructs the center of the cone beautifully and gets progressively less accurate toward the edges, with some unavoidable distortion in the off-center planes. Medical CT increasingly uses heavier iterative reconstruction that costs more computation and buys lower noise. Cone beam trades that away for speed and simplicity.

What you gain is real. The voxels in a CBCT are isotropic, equal in all three dimensions, little cubes rather than the columnar voxels of older CT. That means you can measure and reformat in any plane and trust the geometry, which is exactly what you want for implant planning or localizing an impacted canine against a nerve. You also get that detail at a fraction of the dose and cost of a hospital scan, on a machine that fits in a dental office.

What you give up is contrast, and the culprit is scatter. Because the cone illuminates a large volume at once and the flat panel has no septa between detector rows to reject stray photons, a cone beam scan records an enormous amount of scattered radiation, with a scatter-to-primary ratio that dwarfs the tightly collimated fan beam. Scarfe and Farman name this directly as CBCT’s main disadvantage: the scatter reduces contrast and raises noise. This is not a tuning problem you can fix in software. It is baked into the geometry.

The Hounsfield unit trap

Which brings me back to the implant note. A true Hounsfield unit is a calibrated number. On a medical CT, water is zero and air is negative one thousand by definition, and every gray value maps to a real, reproducible density, which is why radiologists can quote a number and mean something universal by it.

A cone beam scanner cannot give you that. The heavy scatter, the frequently truncated field of view that never captures the whole patient, the noise, and the approximate reconstruction all conspire so that the gray values drift with the machine, the protocol, and even where in the volume the voxel sits. Studies comparing CBCT gray values to true CT Hounsfield units find the correlation positive but weak, with low predictive reliability. Your console may display a number and even label it “HU.” That number is a device-specific gray value wearing a costume. It is fine for a qualitative sense of dense versus soft bone. It is not a calibrated density, and building an implant decision on it as if it were a real Hounsfield unit is trusting a measurement the physics never made.

I did not want to lecture the dentist, so I told him what the scan could honestly support: the bone looked corticated and well structured at the site, the volume for the implant was there, and the gray values were consistent with solid bone, but the specific HU figure the lecture promised was not something cone beam can validly produce. That is a more useful answer than a false number.

Lower dose, and a hard limit on soft tissue

The dose story favors cone beam, and it is one of the genuine reasons the technology spread. For an equivalent maxillofacial region, CBCT generally delivers less dose than a medical CT. Ludlow’s meta-analysis puts average dental cone beam effective doses in the tens to low hundreds of microsieverts depending on field of view, while a medical maxillofacial CT typically runs several times higher. If the question is a bony one that cone beam can answer, it is usually the lower-dose way to answer it.

But that advantage evaporates the moment the question turns to soft tissue, because here cone beam simply cannot compete. The same scatter that wrecks its density values also flattens its soft-tissue contrast. CBCT cannot meaningfully distinguish muscle from tumor, cannot characterize a fascial space infection, cannot take intravenous contrast to light up a fluid collection, and cannot show the TMJ disc. An evidence-based review of cone beam in oral surgery says it directly: CBCT lacks the depiction of soft tissue needed to evaluate pathology, infection, and the joint disc. For those questions the honest tools are medical CT with contrast, which excels at abscesses and cortical erosion and nodal disease, and MRI, which is the modality for soft-tissue extent, marrow invasion, and perineural spread.

Reading the machine for what it is

So the practical map falls out of the physics without much argument. Cone beam is the right instrument for hard-tissue, high-contrast questions: implant sites, impactions and their relationship to the canal, endodontic problems where two-dimensional films are inconclusive, the bony extent of a lesion, the osseous side of the TMJ. It is the wrong instrument, and sometimes a dangerous one, for a suspected malignancy, a soft-tissue mass, a spreading infection, a disc that may be displaced, or any bone-density number you plan to quote to three significant figures.

None of this is a knock on cone beam. I build tools on these volumes and I would not trade the technology away. It is a knock on the assumption hiding in those three shared letters, the assumption that a CBCT is a small cheap CT that does the same job for less. It is not a smaller CT. It is a different camera, tuned hard toward bony detail at low dose and tuned away from soft tissue and calibrated density, and the whole skill of using it well is knowing which questions it was built to answer and refusing to ask it the ones it was not.

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 →