BRENNEN

Industries / Health / Healthcare / Dental

Exams and imaging

Exams and imaging is how a dentist finds disease that cannot be seen from the chair, mostly with X-ray pictures of the teeth and jaws, and how those pictures are read.

7

Milestones achieved, from 1896 to 2026.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1896

    Glass plate in the mouth

    Within weeks of the report of X-rays, the dentist Friedrich Otto Walkhoff wrapped an ordinary photographic glass plate in rubber dam, held it between his teeth and tongue, and lay on the floor through a 25-minute exposure. A 2011 review in Dentomaxillofacial Radiology calls it the first dental radiograph, taken barely 14 days after the discovery was announced. Later that year Walkhoff saw hair loss on the heads of some patients he had X-rayed for 30 minutes at a time. Early operators had no protection: the dentist C. Edmund Kells, another pioneer of dental X-rays, developed a radiation-induced cancer in 1922 and went through 42 operations and several amputations.

    Lessona picture taken many times a day must be judged by the dose it adds up to for the person taking it, not only by what it shows.

    pmc.ncbi.nlm.nih.gov

    Credit

    Friedrich Otto Walkhoff

  2. 1961

    Panoramic X-rays from Finland

    The Finnish dentist Yrjö V. Paatero conceived panoramic radiography in January 1949, and in 1951 took what the Helsinki University Museum calls the world's first panoramic X-ray of a patient's skull. With the engineer Timo Nieminen he built the orthopantomograph, in which the film and the X-ray tube turn around a seated patient and give a right-angled view of every tooth on one film. Commercial production began in 1961.

    blogs.helsinki.fi

    Credit

    Yrjö V. Paatero and Timo Nieminen

  3. 1989

    Sensors replace film

    F. Mouyen and colleagues at the Toulouse Dental School in France described RadioVisioGraphy (RVG), an intraoral X-ray system that captures the image on an electronic light-sensing chip, a charge-coupled device, in place of film. It needed considerably less radiation and showed the picture right after exposure, though at slightly lower resolution than film. A 1991 test of a newer RVG put its resolving power at 11 line pairs per millimeter, against up to 14 for intraoral film.

    ebi.ac.ukebi.ac.uk

    Credit

    F. Mouyen, C. Benz, E. Sonnabend and J. P. Lodter

  4. 1998

    Cone-beam CT for teeth

    P. Mozzo and colleagues at the University Hospital of Verona, Italy, reported a CT machine made for the jaws and face. Instead of the fan-shaped beam of medical CT it uses a cone-shaped beam, and its software builds a 3D volume that can be cut into any slice. It was aimed first at planning dental implants. In their tests on phantoms (stand-ins for a patient), the central dose was about one sixth that of spiral CT, and width measurements were accurate to within 0.8% to 1%.

    ebi.ac.uk

    Credit

    P. Mozzo, C. Procacci, A. Tacconi, P. T. Martini and I. A. Andreis, University Hospital, Verona

  5. 2022

    AI cleared as second reader

    On 4 March 2022 the US Food and Drug Administration (FDA) cleared Pearl's Second Opinion, software that marks suspected decay, faulty filling edges, calculus (hardened plaque) and dark areas at root tips that can signal infection, on bitewings (X-rays of the back teeth taken with the film held between them) and periapical X-rays of patients aged 12 and over. In Pearl's reader study, 25 readers each read 1,005 images without the software and 1,005 with it, and their accuracy improved by a statistically significant margin for all four findings. The FDA's comparison device was Carestream's Logicon Caries Detector, an older computer aid for finding decay.

    accessdata.fda.gov

    Credit

    Pearl, Inc.; reviewed by the FDA

  6. 2024

    Lead aprons no longer advised

    On 1 February 2024 the American Dental Association (ADA) published recommendations, written by an expert panel of its Council on Scientific Affairs with support from FDA medical physicists, that dentists stop putting lead aprons and thyroid collars on patients for dental X-rays, for every patient, including during pregnancy. The shields can block the beam and force retakes. The panel advised narrowing the beam to the area needed (rectangular collimation), using digital sensors and reusing earlier images.

    Lessona shield that spoils the picture can add dose through retakes; dose is cut more reliably at the source of the beam.

    ada.org

    Credit

    ADA Council on Scientific Affairs expert panel, with FDA medical physicists

  7. 2026

    Eight caries AIs compared

    Eduardo Trota Chaves and colleagues at Radboud University Medical Center in Nijmegen, the Netherlands, tested five commercial systems (Second Opinion, CranioCatch, Diagnocat, DIO Inteligência and Align X-ray Insights) and three experimental ones on 200 anonymized bitewings holding 885 filled tooth surfaces, looking for new decay next to fillings. All of them rarely flagged a healthy surface (specificity 0.957 to 0.986) but found only about a third to a half of the lesions (sensitivity 0.327 to 0.487), and none was significantly more accurate than the others.

    Lessoncommercial and experimental systems behaved alike, so each tool has to be checked on the kind of lesion a dentist actually needs it to find.

    ebi.ac.uk

    Credit

    Eduardo Trota Chaves and colleagues, Radboud University Medical Center

Watch for

  1. Dental MRI without radiation

    Dentsply Sirona and Siemens Healthineers say the FDA has cleared MAGNETOM Free.Max Dental Edition, which they call the first dental-dedicated MRI system. MRI makes pictures with a magnet and radio waves instead of X-rays; this one uses a dental coil and software that frame the teeth and jaws without showing the brain. The companies say a clinical trial, described in their own 2025 evaluation report kept on file, showed it can tell active inflammation from healthy and scar tissue and check whether a tooth's pulp is alive. It is meant for hospitals, large clinics and universities.

    Achieved whena peer-reviewed study by a group independent of the makers shows that dental MRI changes diagnosis or treatment compared with X-rays and cone-beam CT, or routine use is reported outside the launch sites.

    biospace.com

    Claimed by

    Dentsply Sirona and Siemens Healthineers, March 2026

Open

  1. 01

    An X-ray order checker

    An open tool that reads a made-up patient record (age, stage of tooth development, new or returning patient, caries risk, symptoms, and the date and type of the last X-rays) and says which X-rays, if any, the ADA and FDA recommendations support, quoting the cell of the recommendations table it used. With it, a public test set of 500 made-up records, each labeled independently by three dentists.

    The brief
    Credit · unclaimed
  2. 02

    Sealed exam for caries AI

    A shared test for dental X-ray AI whose answers no developer can see. Take public dental X-rays that carry no labels yet, such as the 1,571 unlabeled panoramic X-rays released with the DENTEX challenge, have at least four dentists mark decay, deep decay and infection at the root tip on 500 of them, keep those labels sealed on a scoring server, and publish a leaderboard of every system that submits, broken down by kind of lesion.

    The brief
    Credit · unclaimed

Open 01 · the brief

An X-ray order checker

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open tool that reads a made-up patient record (age, stage of tooth development, new or returning patient, caries risk, symptoms, and the date and type of the last X-rays) and says which X-rays, if any, the ADA and FDA recommendations support, quoting the cell of the recommendations table it used. With it, a public test set of 500 made-up records, each labeled independently by three dentists.

Why it is still open

The recommendations are a table by type of visit, age and caries risk, first published by the FDA in 1987 and revised with the ADA in 2004 and 2012. The American Academy of Pediatric Dentistry's 2025 best-practice document, which reprints the table, warns that age or insurance coverage are not on their own reasons to take X-rays and that using insurance payment as a reason may lead to overuse. The ADA's 2024 recommendations ask dentists to order X-rays in moderation and reuse earlier images. A search of journals and the web found no published tool that checks an order against the table and no public test set to score one.

aapd.org

Built with what exists

The table as reprinted by the AAPD; a model built to answer fixed-choice questions in bulk, such as Gemini Decision 1, which returns probabilities for up to 1,000 such questions in one request (see the Note "A thousand questions, one pass"), or a general model such as Claude with the table in its instructions; records generated from templates; a spreadsheet for the three labelers.

  1. First step

    Turn the table into a list of rules, write 50 records that between them hit every cell, and score one model against your own labels.

  2. Done looks like

    On 500 made-up records, the checker agrees with the majority of the three dentists on at least 95%, never advises an X-ray where all three said none was indicated, and every answer quotes its table cell. Records, labels, code and scores are public.

Open 02 · the brief

Sealed exam for caries AI

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A shared test for dental X-ray AI whose answers no developer can see. Take public dental X-rays that carry no labels yet, such as the 1,571 unlabeled panoramic X-rays released with the DENTEX challenge, have at least four dentists mark decay, deep decay and infection at the root tip on 500 of them, keep those labels sealed on a scoring server, and publish a leaderboard of every system that submits, broken down by kind of lesion.

Why it is still open

A call for participation put to the ITU and WHO's Focus Group on AI for Health in September 2022 describes a dental topic group, led from Charité in Berlin, that would benchmark dental AI on undisclosed test data and post a leaderboard; no leaderboard from it was found. DENTEX's test labels are now public, so it can no longer serve as a sealed test. The 2026 Radboud comparison used 200 bitewings from its own clinic, and a 2026 systematic review of 28 studies found only five with external validation and concluded that the evidence "does not support a transferable accuracy benchmark".

itu.inthuggingface.coebi.ac.uk

Built with what exists

The DENTEX images (license CC BY-NC-SA 4.0, on Hugging Face); an open labeling tool such as CVAT; a challenge host such as grand-challenge.org, which ran DENTEX; open baselines such as a small vision model fine-tuned on one graphics chip (see the Note "One GPU to fine-tune"); commercial systems submitted by their makers.

  1. First step

    Ask the DENTEX organizers and the ITU/WHO dental group whether a sealed set already exists, then have three dentists label 50 unlabeled DENTEX images to measure how often they agree.

  2. Done looks like

    500 images with sealed labels on which at least three of four dentists agree; at least five systems scored, including at least two commercial and two open ones; a public leaderboard giving, for each kind of lesion, the share found and the false marks per image; labels released after two years.