BRENNEN

Industries / Health / Healthcare / Dental

Implants

Implant dentistry replaces lost teeth with titanium screws set into the jawbone, which heal tight to the bone and then carry crowns, bridges or dentures.

5

Milestones achieved, from 1965 to 2024.

2

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1965

    Titanium screws in a jaw

    In the 1950s Per-Ingvar Brånemark, studying blood flow and bone healing in rabbits, found that the titanium optical devices he had placed in their leg bones could not be taken out: the bone had grown tight against the metal, which he named osseointegration. In 1965 he placed four titanium implants in the lower jaw of Gösta Larsson of Gothenburg, Sweden, to hold a fixed bridge of teeth, and Larsson kept them until his death some 40 years later.

    dental-tribune.comnobelbiocare.com

    Credit

    Per-Ingvar Brånemark and his team in Gothenburg

  2. 1981

    Fifteen years of results

    Brånemark's team reported in 1981 on every implant placed from 1965 to 1980: 2,768 implants in 410 jaws of 371 patients. In the group treated with the settled method and followed for 5 to 9 years, 81% of upper-jaw and 91% of lower-jaw implants were still stable, and the bone around them shrank 1.5 mm on average during healing and the first year with the bridge, and 0.1 mm a year after that. By Dental Tribune's account, Sweden's National Board of Health and Welfare had approved the system in 1970, Bofors Nobelpharma (now Nobel Biocare) began selling it in 1981, and after Brånemark presented the results at a conference in Toronto in 1982 the FDA soon approved it for the US.

    ebi.ac.ukdental-tribune.com

    Credit

    R. Adell, U. Lekholm, B. Rockler and P.-I. Brånemark

  3. 2016

    Implant disease is common

    Peri-implantitis is inflammation around an implant that destroys the bone holding it. Jan Derks, Tord Berglundh and colleagues at the University of Gothenburg examined 588 people, drawn from a random sample of Sweden's social insurance register, nine years after they had received implants. 45% had peri-implantitis and 14.5% a moderate or severe form, with more than 2 mm of bone lost. Risk was higher in people with a history of gum disease or four or more implants, with certain implant brands, and when the crowns were made by general dentists. The team's earlier paper on the same sample noted that implant results had mostly been measured as survival in small, selected groups at specialist and university clinics.

    LessonAn implant that is still in place can still be failing, and survival counts from selected clinics hid disease that a random national sample found.

    ebi.ac.ukapi.crossref.org

    Credit

    Jan Derks, Tord Berglundh and colleagues, University of Gothenburg

  4. 2016

    A robot arm guides drills

    On 22 December 2016 the FDA cleared the Neocis Guidance System, now sold as Yomi, for dental implant surgery. The surgeon plans the implant on a CT scan. A robot arm then holds the drill and lets the surgeon move it only along the planned path, while a splint attached in the mouth tracks any movement of the head. The arm does not move unless the surgeon pushes the drill, and the summary gives its lateral and depth accuracy as under 1 mm (root mean square). The FDA's October 2026 clearance of the newer Yomi S, for implants anchored in the pterygoid bone behind the upper jaw, keeps the surgeon's hand on the drill.

    accessdata.fda.govaccessdata.fda.gov

    Credit

    Neocis

  5. 2024

    Robots drill on their own

    In China, robots that prepare the implant site on their own, which studies there call autonomous, are in clinical use. At West China Hospital of Stomatology, Muqiao Yu, Yi Man and colleagues compared 71 implants placed with an autonomous robot against 64 placed by hand under dynamic navigation (a screen that tracks the drill), between December 2022 and March 2024. The robot's implants ended a mean 0.50 mm from the plan at the top and 1.62° off in angle, against 1.12 mm and 3.61° with navigation, and drilling took 8.3 minutes instead of 10.6. A 2023 bench study at Shanghai Ninth People's Hospital, on 40 plastic jaw models, also found a task-autonomous robot more precise than navigation.

    ebi.ac.ukebi.ac.uk

    Credit

    Muqiao Yu, Yi Man and colleagues, West China Hospital of Stomatology; Jinyan Chen and colleagues, Shanghai Ninth People's Hospital

Watch for

  1. Autonomous robot leaves China

    BioSpectrum Asia reported on 29 July 2026 that Foundation Healthcare Holdings had launched an autonomous implant robot made by Beijing Yakebot Technology at Farrer Park Medical Centre in Singapore, calling it "Singapore's first autonomous dental implant surgical robot". The report says the robot is the first dental surgical robot to hold both European CE certification and approval through the special review route for new kinds of devices run by China's National Medical Products Administration (NMPA), and it gives no accuracy figures or case counts.

    Achieved whenthe CE certificate can be checked in the EU's EUDAMED database or with its notified body, and a hospital outside mainland China publishes accuracy results for the robot.

    biospectrumasia.com

    Claimed by

    Beijing Yakebot Technology and Foundation Healthcare Holdings, as reported by BioSpectrum Asia, July 2026

  2. AI picks implant positions

    Xinquan Yang and eight co-authors posted RegFreeNet, a network that predicts where an implant should go in a cone-beam CT scan (a 3D dental X-ray), with ImplantFairy, which they describe as a public set of 1,622 scans with implants labeled voxel by voxel. They report state-of-the-art results on ImplantFairy and two other public data sets.

    Achieved whenanother group reproduces the accuracy on scans the authors did not use.

    arxiv.org

    Claimed by

    the authors, in an arXiv preprint, January 2026

Open

  1. 01

    An open implant robot

    A robot drilling system whose code, parts list and test results are all public: a general-purpose six-axis arm carrying a dental implant motor, a camera or optical tracker that follows the jaw, and planning software that turns an implant plan into a drill path. It is tested only on 3D-printed jaws made from public jaw models, with the nerve canal printed as an open channel so that a hit is visible.

    The brief
    Credit · unclaimed
  2. 02

    Planted unsafe implant plans

    A public test set of at least 1,000 virtual implant plans placed in public cone-beam CT scans, half of them safe and half planted with known faults: closer to the inferior alveolar canal (the nerve channel in the lower jaw) than a stated safety margin, into the maxillary sinus, against a neighboring root, or through the jaw's outer wall. Labels are computed from the scans' annotated outlines and a sample is checked by two dentists. A script scores any planning tool or AI model on how many faulty plans it flags and how many safe plans it wrongly flags.

    The brief
    Credit · unclaimed

Open 01 · the brief

An open implant robot

Credit · unclaimed

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

What to build

A robot drilling system whose code, parts list and test results are all public: a general-purpose six-axis arm carrying a dental implant motor, a camera or optical tracker that follows the jaw, and planning software that turns an implant plan into a drill path. It is tested only on 3D-printed jaws made from public jaw models, with the nerve canal printed as an open channel so that a hit is visible.

Why it is still open

The robots that drill on their own are closed commercial products, and the US-cleared Yomi keeps the surgeon's hand on the drill. Research systems built on general-purpose arms are described but not released: Kan and colleagues at Zhejiang University of Technology (2021) positioned a collaborative arm by hand and let it drill on its own into a 3D-printed jaw in a phantom head, and reported mean errors of 1.04 mm at the implant head, 1.56 mm at the tip and 3.74° in angle. Web and GitHub searches in November 2026 found no open-source dental implant robot.

api.crossref.org

Built with what exists

A six-axis arm such as the Universal Robots UR5e or the UFactory xArm 6; ROS 2 and MoveIt for motion planning; 3D Slicer for planning; jaw and canal shapes from the ToothFairy2 data set (CC BY-NC-SA license), printed in resin; scan bodies and a desktop 3D scanner to measure where each implant ended up, with VDING, an open-source program for measuring how far implants land from the plan (Liang and colleagues, PLOS ONE, 2019); a dental implant motor and drills used only on plastic.

  1. First step

    Mount an implant motor on a borrowed arm, print one jaw, drill ten holes to a plan, and measure them with a scanner.

  2. Done looks like

    100 implants placed in 20 printed jaws and measured by someone other than the builders, with mean entry, tip and angle errors published next to the bench figures for a task-autonomous robot (0.46 mm, 0.56 mm and 1.36° at healed sites in the 2023 Shanghai study), and every file released under an open license.

Open 02 · the brief

Planted unsafe implant plans

Credit · unclaimed

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

What to build

A public test set of at least 1,000 virtual implant plans placed in public cone-beam CT scans, half of them safe and half planted with known faults: closer to the inferior alveolar canal (the nerve channel in the lower jaw) than a stated safety margin, into the maxillary sinus, against a neighboring root, or through the jaw's outer wall. Labels are computed from the scans' annotated outlines and a sample is checked by two dentists. A script scores any planning tool or AI model on how many faulty plans it flags and how many safe plans it wrongly flags.

Why it is still open

Public data exist for the parts but not for the check. ToothFairy2, from the University of Modena and Reggio Emilia, holds 480 scans with 42 labeled structures, including the nerve canals, the sinuses and implants, and its 2024 challenge scored the tracing of those structures. ImplantFairy (January 2026) scores where a model would put an implant. Neither tests whether a tool catches a plan that would hit the nerve. A question-and-answer test of AI on cone-beam CT scans, posted to Hugging Face in November 2026 as cbct-doctor-review, asks about teeth, disease and dental work, and has no question on implant plans. Searches of Zenodo, Hugging Face, GitHub and the web in November 2026 found no such benchmark.

ditto.ing.unimore.itarxiv.orghuggingface.co

Built with what exists

ToothFairy2 (CC BY-NC-SA, free account required); 3D Slicer with the open slicerPano extension for implant planning, to place virtual implants; Python with SimpleITK to measure distances; canal-tracing models trained with nnU-Net, the format ToothFairy2 uses; small open vision models such as Liquid AI's LFM3 family, which Liquid says can be fine-tuned on one graphics chip (see the Note "One GPU to fine-tune"), as one kind of tool to test.

  1. First step

    Download ToothFairy2, place 50 virtual implants in 3D Slicer, and publish their measured distances to the canal.

  2. Done looks like

    1,000 plans and the scoring script public; two dentists agree with the computed label on at least 95% of a 200-plan sample; and published scores for at least five tools, including one commercial planning program.