Industries / Health / Healthcare / Dental
Crowns and lab work
Making the crowns, bridges, inlays and dentures that cover or replace teeth: shaping the tooth to receive them, recording its shape, and the dental lab or machine that builds the piece.
6
Milestones achieved, from 1889 to 2025.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1889
An all-porcelain crown
A crown is a cap that covers a damaged tooth. Charles H. Land patented an all-porcelain "jacket" crown in 1889. Gregg Helvey's history of dental ceramics says jacket crowns were not known for strength, because of internal microcracking, yet they were used widely until the 1950s.
Credit
Charles H. Land
1962
Porcelain fused to metal
Porcelain baked onto a thin metal cap gave a crown the strength of metal under a tooth-colored surface. Morris Weinstein, Sigmund Katz and Abraham B. Weinstein filed a patent for "fused porcelain-to-metal teeth" on 15 October 1959, and it was granted on 11 September 1962, after Morris Weinstein's death. Helvey dates the crown to the late 1950s and notes its trade-off: fewer porcelain failures, but an opaque layer that hides the metal made it look less natural.
Credit
Abraham B. Weinstein, Morris Weinstein and Sigmund Katz
1965
Alumina strengthens porcelain
John W. McLean and T. Harry Hughes, in Britain, reinforced dental porcelain with alumina, a hard industrial ceramic, to make a stronger core for all-porcelain crowns, and reported it in the British Dental Journal in 1965. Helvey calls this the start of a revival of all-ceramic restorations.
Credit
John W. McLean and T. Harry Hughes
1985
Inlays milled at the chair
An inlay is a small piece made outside the mouth to fill a cavity. On 19 September 1985, at the dental institute of the University of Zurich, Werner Mörmann and the electrical engineer Marco Brandestini made a ceramic inlay with CEREC, a system they built from a 3D camera that records the prepared tooth in the mouth and a milling unit that carves the piece from a ceramic block. The piece can be made and fitted at one visit, which the trade calls chairside. A 2016 press release marking Mörmann's 75th birthday put the count at more than 35 million CEREC restorations and about 38,000 active users.
Credit
Werner H. Mörmann and Marco Brandestini, University of Zurich
2008
Lead scare over imported crowns
In February 2008 the US television station WBNS-TV reported lead in dental work made in China: a bridge removed from a patient held 160 parts per million (ppm), and one of eight porcelain-fused-to-metal crowns it ordered from four Chinese labs held 210 ppm. The American Dental Association then analyzed 44 porcelain powders and 102 finished crowns from US and foreign labs, and found trace lead that occurs naturally in porcelain but no evidence that it leached into the mouth. The coverage revived a debate over requiring dental labs to be certified and registered.
LessonPatients rarely knew where their crowns were made, and measurement, not the country of origin, settled the safety question.
Credit
WBNS-TV; American Dental Association
2025
AI crowns match technicians
Researchers at University Hospital Regensburg and OTH Regensburg, in Germany, designed crowns for one resin lower first molar in five ways, including two AI design tools, 3Shape's Automate and Imagoworks' Dentbird, and designs refined by a dental technician, then milled and tested them. The AI designs matched the technician's in fit at the margin, in function and in the force needed to break them. No design, human or AI, touched all the opposing teeth it should have, mostly because of errors on one cusp. The study appeared in Dental Materials in September 2025.
Credit
Alexander Broll, Martin Rosentritt and colleagues
Watch for
A robot inside the mouth
Engineers at the University of Basel built MIR, a robot about the size of a wine cork meant to sit in the mouth and cut a tooth for a crown from a digital plan. On model teeth of resin and of a ceramic about as hard as enamel, it placed its burs with an error under 0.2 mm without position sensors, and drilling forces stayed below 5 newtons. The team says more development is needed before it can be used on patients.
Achieved whena group outside the team measures its accuracy on the same kind of model teeth, or a clinical study reports crown preparations it made in patients.
Claimed by
Yukiko Tomooka, Georg Rauter and colleagues, University of Basel, with the University of Zurich, the University of Bern and Camlog, June 2026
Open
01
One test for prep robots
A public test kit for machines that cut teeth to take crowns: files for printing a standard model jaw, a list of stock typodont teeth to buy, a target preparation for each tooth agreed by two prosthodontists (dentists who specialize in crowns and dentures), and an open script that lines up a scan of the cut tooth with its target and reports the error at the margin, on the side walls and on the biting surface. Then score dental students, a robot that follows a planned path and a robot taught by demonstration on the same kit.
The briefCredit · unclaimed02
An open crown-design benchmark
A public benchmark of at least 30 typodont preparations covering front and back teeth, each scanned with its neighbors and the opposing teeth, plus a scan of the original uncut tooth as the reference shape. At least three AI design tools and two dental technicians each design a crown for every preparation. Every design is milled in one material and measured for the gap at its margin, its contacts with neighboring teeth and its contacts with the opposing teeth. The files and the scoring script are published.
The briefCredit · unclaimed
Open 01 · the brief
One test for prep robots
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public test kit for machines that cut teeth to take crowns: files for printing a standard model jaw, a list of stock typodont teeth to buy, a target preparation for each tooth agreed by two prosthodontists (dentists who specialize in crowns and dentures), and an open script that lines up a scan of the cut tooth with its target and reports the error at the margin, on the side walls and on the biting surface. Then score dental students, a robot that follows a planned path and a robot taught by demonstration on the same kit.
Why it is still open
Each team tests its robot its own way. Researchers at Beihang University reported in April 2026 that their robot arm prepared crowns on 11 standardized models with an average error of 0.18 mm, against 0.33 mm for 11 done with step-by-step reduction guides, on models of their own. The University of Basel tested its MIR robot on its own resin and ceramic teeth. The claims about Perceptive's crown robot in 2024 came from its chief executive and an investor, with no independent check reported. Scoring software exists for teaching: Dentsply Sirona's prepCheck compares a student's scanned preparation with a master preparation or measures its taper, reduction and margins, as a 2020 study at the University of Groningen describes, but it is closed software and each school sets its own targets. A 2021 course project at Worcester Polytechnic Institute planned a robot's crown preparation path in MATLAB simulation. A search of Hugging Face's datasets found no public tooth-preparation data.
buaa.globalimpact.cnspectrum.ieee.orgpmc.ncbi.nlm.nih.govgithub.com
Built with what exists
Typodont teeth of the kind sold to dental schools; a desktop dental lab scanner; the open libraries Open3D or CloudCompare for lining up scans and mapping distances; an electric micromotor handpiece with diamond burs; an industrial arm, or a low-cost one taught with LeRobot (see the Note "No download") or Runway's open Praxis-1 robot model (see the Note "Open robot model weights").
First step
Buy 20 identical typodont molars, scan one, draw its target preparation, and write the alignment script in Open3D.
Done looks like
The kit published under an open license, with scans and scores for at least 20 preparations each by students, a scripted robot and a taught robot, and a second lab that reruns the script on the published scans and gets the same numbers to within 0.01 mm.
Open 02 · the brief
An open crown-design benchmark
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public benchmark of at least 30 typodont preparations covering front and back teeth, each scanned with its neighbors and the opposing teeth, plus a scan of the original uncut tooth as the reference shape. At least three AI design tools and two dental technicians each design a crown for every preparation. Every design is milled in one material and measured for the gap at its margin, its contacts with neighboring teeth and its contacts with the opposing teeth. The files and the scoring script are published.
Why it is still open
The Regensburg comparison of AI and technician designs used a single lower molar and found that no design touched all the opposing teeth it should have. Data for training crown-design AI is starting to appear: CrownFusion, presented at MICCAI 2026, shares code that generates a crown from scans of the preparation and neighboring teeth, and its author wrote on 17 November 2026 that the data set is ready but its release is delayed; ToothCraft (2026) completes damaged crowns with a model trained on synthetic data. Both generate shapes in software, and neither repository describes milling and fitting the results. A search of Hugging Face's datasets for crown, preparation and restoration data found none.
Built with what exists
Typodont teeth; a desktop lab scanner; AI design tools such as 3Shape Automate and Imagoworks Dentbird, with exocad as the conventional baseline and the open-source DentCraft, released in October 2026 and still in early development, as an open one; a lab milling machine and resin or zirconia blocks; a digital microscope or the silicone replica method for margin gaps; Open3D for comparing shapes.
First step
Prepare and scan five typodont teeth, one of each tooth type, and run them through two AI design tools.
Done looks like
At least 150 milled crowns measured by someone who does not know which tool designed each, with the mean margin gap in micrometers and the share of planned contacts reached, per tool and per tooth type, and every file public.