September 11, 2026 The Mouth on the Screen: Diagnosing Childhood Decay From a 3D Intraoral Scan
For as long as decay has been diagnosed in a small child’s mouth, the method has been almost unchanged: a clinician, a mirror, a good light, and a patient willing to sit still and open wide long enough to be examined. Everything about that ritual depends on the child’s cooperation and the examiner’s presence in the room at that exact moment. A 2025 Australian study, led by Bree Jones at the Murdoch Children’s Research Institute and the University of Melbourne, quietly proposed a different arrangement. Capture a colour three-dimensional intraoral scan of the child’s teeth, and let a clinician read the decay from the model on a screen – and the result agreed strongly with the standard chairside visual examination, across the full range from the earliest white-spot lesions through to frank cavitation. The finding is easy to under-read as just another scanner study. Its real weight is this: the scan is beginning to carry the diagnosis, not merely the impression.
That is a genuine shift in what an image is for. The intraoral scanner earned its place in dentistry as an impression-replacement – a way to capture arch geometry for a crown, an aligner, or a full-arch restoration without a tray of material. In that role it is a means to a model, and the model is a means to a device. Reading caries off the scan asks the image to become something else entirely: the diagnostic object itself, the thing the clinician judges instead of the mouth. When the picture inherits the diagnosis, every property of the picture stops being cosmetic and starts being clinical.
When the Image Becomes the Patient
The exam chair has one great advantage that is easy to forget: the clinician is looking at the actual tooth, with stereoscopic vision, an adjustable light, an air syringe to dry a suspicious surface, and the freedom to change the angle at will. To match that from a captured model, the image has to reproduce enough of the tooth that nothing diagnostic is lost in the translation. The study’s central result – strong agreement between on-screen reading and the in-mouth exam – is really a statement about the fidelity of the capture: for these surface lesions, the colour 3D model contained enough of the truth that a trained reader reached the same verdict without the patient present. That is the whole game. The diagnosis is only as trustworthy as the image is honest.
What the Image Has to Contain
Visual caries detection is, at bottom, the reading of colour and texture. The vocabulary is tonal: the chalky, matte white of early demineralisation; the dark, glassy brown of an arrested lesion; the shadowed pit of frank cavitation. A scan meant to support that reading must therefore capture colour faithfully – not the flattering, contrast-boosted colour that makes a marketing render look clean, but accurate tone, because an over-processed image can erase a subtle white-spot or invent one that is not there. It needs spatial resolution fine enough to resolve surface texture and the edge of a micro-cavitation, and accurate surface geometry so that a lesion’s depth and contour read correctly in three dimensions rather than flattening into an ambiguous smudge. These are the same disciplines that govern the accuracy craft of intraoral scanning for restorative work – faithful geometry, honest capture – now asked to serve diagnosis, where the tolerance for error is different in kind.
The Enemy Is Capture, Not the Concept
Anyone who has scanned a young child knows the failure modes, and they are all failures of capture rather than of the idea. Saliva pools and bubbles that read as spurious texture; a tongue or cheek that drifts across the field and occludes a surface; motion blur and stitching errors from a patient who will not – cannot – hold still. Each of these corrupts the very surfaces a caries reading depends on. The scan’s promise of being child-friendly is real, but it is conditional: the capture is fast and non-threatening, yet a rushed, wet, or motion-degraded scan simply moves the difficulty from the exam to the image. This is the familiar law of the field – the diagnosis inherits every flaw in the picture – and it is why the honest reading of a scan begins with an honest scan.
The Quiet Revolution: Decoupling Capture From Reading
The deepest consequence of a diagnosis that lives in an image is that capture and interpretation no longer have to happen at the same time or in the same place. A visual exam is synchronous by definition: examiner and child, together, now. A scan is asynchronous. It can be captured in a few gentle passes – potentially by less specialised staff, in a school, a community clinic, or a rural setting – and read later by an expert who may be hundreds of kilometres away. Radiation-free and non-contact, it lowers the threshold for screening children who might never reach a specialist chair. That is the teledentistry argument, and this study gives it something it usually lacks: evidence that the read off the model is close enough to the read off the mouth to be worth trusting. The image becomes a portable, storable proxy for the patient.
Where the Scan Stops and the Radiograph Begins
Honesty about what the scan can do requires equal honesty about what it cannot. A surface scan sees surfaces. It reads occlusal and accessible smooth-surface lesions – exactly the early childhood decay the study targeted – but it is blind to what hides between the teeth and beneath the enamel, where interproximal and dentinal caries do their quiet damage. That territory still belongs to the radiograph. The bitewing remains the instrument that sees through the tooth rather than across it, and the way AI now reads a bitewing for caries and bone loss is a separate diagnostic conversation entirely. The right framing is not scan versus radiograph but scan and radiograph: the colour model owning the visible surfaces, the radiograph owning the hidden interfaces. AI is beginning to sit over both, flagging suspicious surfaces on the model much as it has learned to flag shadows on the film – part of the broader arc from mere detection toward assisted diagnostic reporting.
Future Developments
If the scan can be trusted to carry a surface-caries diagnosis, the natural next step is to trust it to carry that diagnosis through time. A young child scanned at every recall would accumulate not a folder of disconnected snapshots but a continuous, radiation-free, three-dimensional record of the same dentition – a longitudinal gallery in which a white-spot lesion could be watched to see whether it arrests or advances, its colour and contour compared honestly against its own past rather than against memory. That record could travel with the child between clinicians, be re-read as criteria evolve, and be surfaced by software that highlights only what has changed. The condition, always, is the image. A scan is only diagnostic if its colour is true, its surface faithful, and its capture clean enough that nothing meaningful was lost between the mouth and the model. The study’s real lesson is not that the scanner replaces the examiner, but that when we ask a picture to stand in for a patient, the craft of making that picture well stops being an aesthetic nicety and becomes the diagnosis itself.
Sources & further reading:
No Comments