How CT Scans Reveal What's Inside Fossil Skulls

A fossil skull looks solid from the outside, but its most valuable information often sits sealed inside it, in cavities that once held a brain, inner ear canals, and looping nasal passages. Cutting a skull open to look would destroy a specimen that may be the only one of its kind on Earth. Computed tomography (CT) solves that problem by using X-rays to build a three-dimensional digital model of everything inside the bone, without removing a single gram of it. That technique has let researchers reconstruct dinosaur brain shape, map inner ear structure tied to balance and hearing, and trace nasal passages linked to smell and vocalization, all from specimens that remain physically untouched.

What a CT scan actually does to a fossil

Computed tomography — an imaging method that reconstructs a 3D object from many 2D X-ray cross-sections — works by rotating an X-ray source and detector around a specimen while it captures dozens to thousands of individual radiographic slices from different angles. Software then stacks and cross-references those slices, using differences in how densely each point in the specimen absorbs X-rays to reconstruct a full three-dimensional volume. Denser material blocks more X-rays and shows up brighter in the reconstruction; less dense material lets more radiation through and appears darker.

Fossilized bone and the surrounding rock matrix that entombs it usually differ in density, especially when minerals have replaced the original bone tissue during fossilization, a process called permineralization. That density contrast is what lets the software digitally separate bone from rock, letting a researcher isolate the skull, delete the surrounding matrix in software, and rotate a clean 3D model of just the specimen on a screen. It is the same physical principle behind a hospital CT scanner, applied to a specimen that took millions of years to form and is usually orders of magnitude denser than living tissue.

The output isn't a photograph. It's a volumetric dataset, essentially a stack of digital slices a fraction of a millimeter apart, that can be resliced along any plane, measured precisely, 3D-printed, or shared as a digital file with any lab in the world. That last part matters more than it sounds. A cast or a physical loan of a rare fossil requires shipping an irreplaceable object; a CT dataset is an email attachment.

Researchers scanning the Field Museum's *Tyrannosaurus rex* specimen "Sue" (FMNH PR 2081) found olfactory bulb proportions consistent with a strong sense of smell, part of a broader finding that tyrannosaur olfactory bulb size was comparable to or larger than that of many modern carnivorous birds and mammals, based on CT-derived endocast measurements (Brochu, 2003, *Society of Vertebrate Paleontology Memoir*; Witmer & Ridgely, 2009, *The Anatomical Record*).

Why does CT scanning matter more for fossils than for other objects?

Most dinosaur species are known from only one to a handful of specimens worldwide, sometimes just a single skull. Any destructive technique risks losing anatomical information permanently, with no second specimen to check against. CT scanning removes that risk entirely by building an equally detailed digital record while leaving the physical fossil completely intact for future study.

From CT data to a brain: how endocasts get made

A brain endocast — a cast, physical or digital, of the internal cavity that once held a brain — is one of the main products of CT-scanning a skull. Dinosaur brains themselves never fossilize; soft neural tissue decays long before mineralization can begin. What sometimes survives is the bony braincase that surrounded it, and the shape of that internal cavity roughly mirrors the shape of the brain and major nerves that once filled it.

Researchers segment the CT data, digitally tracing the boundary between braincase bone and the empty cavity inside, slice by slice, until software can generate a solid 3D model of just that interior space. The resulting endocast approximates overall brain shape, relative size of major regions like the olfactory bulbs and cerebral hemispheres, and the paths of major cranial nerves and blood vessels that left channels in the surrounding bone.

This method has produced some of the most-cited paleoneurology work of the past 25 years. Christopher Brochu's 2003 monograph on Sue, published as a Society of Vertebrate Paleontology memoir, used high-resolution CT to describe her skull and braincase in more anatomical detail than any Tyrannosaurus specimen before it. A follow-up study by Lawrence Witmer and Ryan Ridgely, published in 2009 in The Anatomical Record, CT-scanned multiple tyrannosaur braincases, including Tyrannosaurus rex, and used the resulting endocasts to compare olfactory bulb size, inner ear geometry, and inferred sensory capability across species (Witmer & Ridgely, 2009).

What does an endocast actually tell researchers about behavior?

An endocast reveals relative proportions, not neuron counts or intelligence in any direct sense. A large olfactory bulb region relative to total brain size suggests a keener reliance on smell; a well-developed inner ear labyrinth oriented a certain way suggests specific head postures or balance demands during locomotion. Researchers treat these as informed inferences, not literal readouts of thought.

The inner ear, balance, and how dinosaurs held their heads

The vestibular system — the inner ear structures responsible for sensing balance, head orientation, and motion — leaves a distinct three-dimensional trace inside skull bone called the bony labyrinth, and CT scanning is essentially the only practical way to study it in a fossil. The labyrinth consists of three semicircular canals arranged roughly perpendicular to each other, each one sensing rotational movement in a different plane.

The orientation of the labyrinth relative to the rest of the skull correlates, in living animals, with habitual head posture: how an animal held its head when its horizontal semicircular canal sat level with the ground, which is the orientation the brain treats as "forward-looking" alertness. Witmer and Ridgely's 2009 CT study of tyrannosaur braincases used this relationship to model probable head carriage angle in life, an inference that would be essentially unobtainable from external bone examination alone (Witmer & Ridgely, 2009).

Think of the bony labyrinth as a sealed spirit level built into the skull. A carpenter's level shows when a surface is flat by how a bubble settles inside a sealed tube; the semicircular canals work on a related principle, using fluid movement inside sealed loops to register orientation and rotation. CT scanning is what lets researchers read that fossil "level" without breaking it open, tracing the empty canal spaces in three dimensions from density data alone.

Because the labyrinth is often only a few millimeters across in even a large dinosaur skull, this kind of analysis demands genuinely high scan resolution, not just any CT scan. Coarser hospital-grade scanners built for human diagnostic imaging can miss fine canal detail entirely, which is part of why paleontology has increasingly turned to industrial and synchrotron-based scanning for the smallest structures.

Sinuses, nasal passages, and why some dinosaur heads sound like instruments

CT scanning has also mapped extensive, often unexpectedly elaborate sinus and nasal passage systems inside dinosaur skulls, cavities that have no analog in most living reptiles and were essentially invisible before this technology. In hadrosaurs, the duck-billed dinosaurs known for their hollow cranial crests, CT scans of species like Parasaurolophus revealed nasal passages that loop back through the crest itself before reaching the throat, rather than running in a short straight line as in most vertebrates.

Researchers studying these looped passages, building on decades of prior work including Witmer's broader research program on dinosaur soft-tissue reconstruction, have proposed that the elongated nasal tract functioned partly as a resonating chamber, shaping vocalizations the way the looped tubing of a brass instrument shapes sound before it exits the bell. That's an inference drawn from the geometry the scans reveal, since no fossil preserves actual sound, but the anatomical case for some acoustic function is considered reasonably strong given how specifically shaped and species-variable the crest passages are.

Sinus systems elsewhere in the skull, including around the eye socket and jaw, have also shown up clearly in CT data where they were previously undetectable from the outside. These air-filled spaces likely served multiple roles across different species, from reducing skull weight to possibly assisting in thermoregulation, and mapping their full extent required exactly the kind of internal, non-destructive view that only tomographic imaging provides.

Beyond hospital CT: how synchrotron and industrial scanning push resolution further

Standard medical CT scanners, the same category of machine used for a human patient, work well for many fossil skulls but hit resolution and density limits with very small structures or extremely dense mineral replacement. Paleontology labs increasingly turn to two more specialized tools for the hardest cases: industrial CT scanners built for denser materials like metal parts, and synchrotron-based microtomography, which uses an extremely intense, highly focused X-ray beam generated by a particle accelerator facility.

Synchrotron microtomography can resolve structural detail at the scale of micrometers rather than the sub-millimeter range typical of hospital CT, which matters enormously for tiny embryonic fossils, individual tooth microstructure, or the finest canals of an inner ear labyrinth. The European Synchrotron Radiation Facility (ESRF) in Grenoble, France, has scanned fossil material including dinosaur embryos and skull fragments at resolutions in the range of roughly 5 to 13 micrometers, fine enough to trace individual growth lines inside fossil teeth (ESRF; Tafforeau et al., published research using ESRF beamlines). That's a level of detail no hospital scanner can reach.

The tradeoff is access and scale. Synchrotron facilities are few, scanning time is scarce and competitively allocated, and the specimen size that fits inside a synchrotron beamline is often smaller than what an industrial CT scanner can handle. Most routine fossil skull work, including large theropod braincases, still relies on industrial or high-resolution hospital-grade CT, reserving synchrotron time for the smallest or most demanding structures.

Imaging technique	Typical resolution	Best suited for	Key limitation
Hospital-grade medical CT	~0.5–1 mm	Full skulls, large braincases, general endocast shape	Misses very fine canal or tooth microstructure
Industrial CT scanning	~0.05–0.3 mm	Dense fossils, large specimens, routine paleoneurology	Slower scans, high equipment cost
Synchrotron microtomography	~1–15 micrometers	Tiny embryos, tooth growth lines, fine labyrinth detail	Very limited facility access, small specimen size
Laser surface scanning	Surface only, sub-millimeter	External morphology, digitizing exposed bone surfaces	Cannot see anything beneath the surface

Each technique answers a different question: surface scanning documents what's visible, while CT and synchrotron methods are the only ways to see structure sealed inside solid bone.

From raw slices to a manipulable model: the segmentation workflow

A CT scanner's raw output is not, by itself, a usable 3D model. It is a stack of grayscale cross-sectional images, sometimes thousands of them, each one showing density values as shades of gray with no labels attached to say "this pixel is bone" versus "this pixel is rock." Turning that stack into something a researcher can rotate, measure, and publish requires a separate step called segmentation, using specialized software such as Mimics, Avizo, VG Studio Max, or the open-source program 3D Slicer.

Segmentation works slice by slice. A researcher, or increasingly a semi-automated algorithm working under human supervision, traces the boundary of a structure of interest, the outer skull surface, the brain cavity, a single nasal passage, on each two-dimensional slice in the stack. Basic thresholding tools can separate regions automatically when density contrast is strong and clean, essentially telling the software "everything brighter than this value is bone, everything darker is rock or air." When contrast is weak or a specimen has been crushed, cracked, or partially replaced by minerals similar in density to the original bone, that automated approach breaks down, and a trained technician has to segment the boundary by eye instead, using subtle textural cues or comparison with better-preserved parts of the same specimen.

Once every slice is segmented, the software stitches the traced boundaries into a continuous 3D surface mesh, the kind of solid, rotatable digital object seen in published paleoneurology figures. That mesh can then be measured for volume, exported for 3D printing, digitally "unbroken" if the original fossil was crushed or distorted, and even used to test hypotheses about biomechanics through digital modeling. A crushed or taphonomically deformed skull, one flattened or twisted during burial, can sometimes be digitally retrodeformed, essentially reversing the distortion in software using the mirrored geometry of the less-damaged side, a correction that would be physically impossible to perform on the actual bone without breaking it further.

This workflow is also where most of the labor in a CT-based paleoneurology study actually happens. The scan itself might take hours; segmenting a complex braincase with dozens of internal cavities and nerve canals down to a publication-ready endocast can take a researcher weeks or months of slice-by-slice tracing, particularly for structures with poor density contrast against surrounding rock.

A case study: what CT found that opening the fossil could not

The clearest demonstration of what non-destructive CT scanning can recover, versus what any physical preparation technique would have destroyed, comes from dinosaur eggs rather than skulls, though the underlying logic is identical. In 2021, researchers led by Fion Waisum Ma described "Baby Yingliang," an oviraptorosaur embryo fossil from Late Cretaceous rock in Ganzhou, China, published in the journal iScience. The specimen had already been partly exposed by a natural crack in the eggshell, but the rest of the embryo's skeleton, including its skull and spine, remained fully encased in rock and shell.

CT scanning let the team reconstruct the embryo's posture inside the intact portion of the egg without breaking any more shell away, revealing that the animal was curled with its head tucked beneath one wing-like forelimb, a posture known as "tucking" that is common in modern bird embryos immediately before hatching but had never before been documented in a non-avian dinosaur (Ma et al., 2021, iScience). Physical preparation, chipping the surrounding shell and matrix away with a needle or air scribe, would have risked destroying exactly the delicate, curled arrangement of bones that made the discovery significant in the first place. CT imaging recovered that posture intact, encased shell and all.

A comparable logic applies to pathology studies on skulls. CT scans of Sue's lower jaw, which carries several smooth-edged holes that puzzled researchers for years, let scientists measure the internal shape and depth of those lesions from the inside out, without drilling into or sectioning the actual bone to see how deep the damage ran. That internal geometry became central evidence in the debate over what caused the holes: an early study proposed a parasitic infection similar to avian trichomoniasis, based partly on lesion shape, while later re-examination of the internal CT geometry argued the pattern was a poor match for that specific parasite and more likely reflects a different cause, possibly related to combat injuries, with the exact cause still unresolved (Wolff et al., 2009, PLOS ONE; Field Museum research summary on later re-examination). Either way, the internal CT data, not surface inspection alone, is what let researchers argue the case at all.

Why this changed what paleontology can know

Before CT scanning became routinely accessible to paleontology labs through the 1990s and 2000s, questions about brain shape, inner ear orientation, and internal sinus anatomy in dinosaurs were largely unanswerable on any specimen a museum wasn't willing to section. Physical sectioning, effectively cutting the skull to expose an internal surface, was sometimes attempted historically on lower-value or duplicate material, but was never an option for something like a one-of-a-kind, near-complete Tyrannosaurus skull.

That access barrier meant entire categories of biological inference, like relative reliance on smell versus vision, or probable head posture during normal locomotion, simply sat outside what science could test for most species. CT scanning opened that door for essentially every reasonably well-preserved skull a museum was willing to loan for imaging time.

The digital nature of the output changed collaboration too. A CT dataset segmented into an endocast can be measured, compared statistically against other species' endocasts, and published with the full underlying data available for other researchers to check, all without anyone needing physical access to the original bone. Witmer's lab at Ohio University, for example, has made numerous digital endocast models publicly viewable, letting anyone examine the same reconstructed brain cavity the original researchers studied.

What CT scanning still can't tell researchers

An endocast is a cast of a cavity, not a cast of the brain itself. In most vertebrates, especially reptiles and their relatives, the actual brain doesn't fill the entire cranial cavity the way a mammal brain typically does; connective tissue, blood vessels, and fluid occupy some of that space too. That means endocast shape is a reasonable but imperfect proxy for true brain shape, and researchers are careful to describe endocast-based conclusions with appropriate caveats rather than treating them as a literal brain scan.

Behavioral inferences drawn from sensory anatomy, like "a large olfactory bulb region means X hunted primarily by smell," are also probabilistic arguments built from comparative anatomy across many living and extinct species, not direct proof for any single individual. Two dinosaurs with similar endocast proportions could plausibly have used those senses somewhat differently depending on ecology, habitat, and behavior that fossils alone can't fully capture. Reputable paleoneurology papers, including Witmer and Ridgely's tyrannosaur work, generally frame their conclusions this way, as the best current inference from the available anatomical evidence rather than a settled fact.

Scan quality itself has practical limits too. Very low density contrast between fossil bone and its surrounding matrix, common in some sedimentary preservation types, can leave the boundary between bone and rock ambiguous even at high resolution, forcing researchers to segment data by hand using other anatomical cues rather than relying on density alone.

The takeaway

CT scanning turned solid fossil skulls into readable, shareable digital objects, letting paleontologists reconstruct brain shape, trace inner ear canals tied to balance, and map nasal passages linked to smell and sound, all without cutting into bone that may be the only example of its species on the planet. The CT-based work on Tyrannosaurus rex specimens like Sue, by researchers including Christopher Brochu, Lawrence Witmer, and Ryan Ridgely, remains some of the clearest evidence of how a technology built for human hospitals ended up reshaping what science can know about a 67-million-year-old animal's senses, all from the inside of a skull no one ever had to open.

---

Read next:

• How Paleontologists Date Dinosaur Fossils
• How Amber Preserves Ancient Ecosystems
• How Dinosaur Footprints Preserve Behavior

FAQ

Q: How does a CT scanner see inside a solid fossil skull?

It rotates an X-ray source and detector around the specimen, capturing many individual X-ray slices from different angles, then reconstructs those slices into a three-dimensional digital model using density differences between fossilized bone and the surrounding rock matrix to tell the two apart.

Q: What did CT scans reveal about the Tyrannosaurus rex brain?

CT-based endocast studies, including work on the Field Museum's "Sue" specimen and Witmer and Ridgely's 2009 comparative tyrannosaur study, found proportionally large olfactory bulb regions suggesting a strong sense of smell, along with inner ear geometry consistent with specific head posture and balance demands, all inferred without physically opening the skull.

Q: Why is CT scanning preferred over physically cutting open a fossil?

Because cutting a fossil open permanently destroys it, and most dinosaur species are known from only one or a few specimens worldwide with no duplicate to fall back on. CT scanning produces an equally detailed digital record while leaving the original fossil fully intact for future researchers and future imaging technology.

Q: Can CT scanning fail on certain types of fossils?

Yes. Fossils with very little density contrast between bone and surrounding rock, or extremely fine structures like inner ear canals that fall below a scanner's resolution, can produce ambiguous or incomplete results, which is part of why some labs turn to higher-resolution synchrotron microtomography for the smallest anatomical details.

Q: What is a brain endocast and how accurate is it?

A brain endocast is a cast, physical or digital, of the internal cavity that once held a dinosaur's brain, built from segmented CT data. It approximates overall brain shape and regional proportions reasonably well, but it isn't a literal brain scan, since connective tissue and fluid also occupied part of that cavity in life, so researchers treat endocast-based conclusions as informed inferences rather than exact measurements.

Q: What is segmentation, and why does it take so long?

Segmentation is the process of digitally tracing the boundary between bone, rock matrix, and empty cavity on every individual slice of a CT scan, using software like Mimics or Avizo, so the computer can stitch those traced outlines into a solid 3D model. It takes so long because weak density contrast, cracks, or mineral replacement often force a technician to trace boundaries by eye rather than relying on automated thresholding, and a complex braincase can involve thousands of slices.

Q: Has CT scanning ever revealed something inside a fossil that physical preparation couldn't have shown?

Yes. A well-documented example is "Baby Yingliang," an oviraptorosaur embryo fossil described in 2021 in the journal iScience, where CT scanning revealed the embryo curled inside its intact egg in a bird-like "tucking" posture just before hatching, a delicate arrangement of bones that physically chipping away the surrounding shell would likely have destroyed before it could be documented.