Overjet Dental Assist and K210187: what the FDA's 2021 clearance lets an AI bone-level reader do

What it says#

The FDA's 510(k) record lists Overjet Dental Assist, applicant Overjet, Inc. of Boston, received January 25, 2021 and decided May 19, 2021: substantially equivalent, under regulation 21 CFR 892.2050 (medical image management and processing system), product code LLZ, Class II, reviewed by the Radiology panel as a traditional 510(k) with no third-party review. The clearance letter, signed for the director of the Division of Radiological Health, says the device may be marketed "for the indications for use stated in the enclosure," which is the whole point of a 510(k): the indication is the permission.

That indication reads, in full: "Overjet Dental Assist is a radiological semi-automated image processing software device intended to aid dental professionals in the measurements of mesial and distal bone levels associated with each tooth from bitewing and periapical radiographs. It should not be used in-lieu of full patient evaluation or solely relied upon to make or confirm a diagnosis. The system is to be used by trained professionals including, but not limited to, dentists and dental hygienists." The intended patient population is "patients living in the United States, who are 22 years old or older, and that do not have any remaining primary teeth," and the form adds that "Overjet has not evaluated the performance of the device on primary dentition." The box checked is prescription use.

Overjet's 510(k) summary, dated April 12, 2021, describes the product as "a cloud native Software as a Medical Device that allows users to automate the measurement of interproximal bone levels for bitewing and periapical radiographs, review associated radiographs, view annotations, modify annotations." The predicate was EZOrtho v1.0 from Ewoosoft (K192888), a locally installed orthodontic image-analysis program; the summary argues both are software that annotates radiographs for a dentist's diagnostic process, with the differences being cloud versus local installation, the interface, and EZOrtho's extra features that Dental Assist does not need.

The testing section is what a clinician should read. Software was treated as a "moderate" level of concern. Bench testing scored precision and recall against labeled keypoints on 2,234 bitewing and 6,543 periapical radiographs (average precision 82.3 and 83.0 percent, average recall 89.5 and 90.0 percent). Clinical testing used retrospective data: 65 bitewing and 96 periapical radiographs from 63 subjects, 73 percent female, ages from under 30 to over 70. Three U.S. licensed dentists labeled interproximal bone levels in Dexis Webview, two U.S. dental radiologists adjudicated, and the software was compared to that ground truth against prespecified criteria of more than 85 percent sensitivity and specificity and a mean absolute difference under 1.5 mm. Results: on bitewings, 98.7 percent sensitivity, 95.0 percent specificity and a 0.307 mm mean absolute difference; on periapicals, 88.94 percent, 95.96 percent and 0.353 mm; periapical root length, 90.9 percent, 97.47 percent and 0.567 mm. The summary notes the periapical bone-level result met the performance criteria but not the 0.05 alpha (it was 0.083), that 33 percent of false negatives involved anatomical overlap obscuring the CEJ, and that the overall 88 percent sensitivity "was consistent with the performance of the three ground truth dentists." No adverse effects were recorded; no treatment decisions were made.

Overjet's release the next day, May 20, 2021, headlined the clearance "In a Dental Industry First," said the software "supports dental professionals in measuring mesial and distal bone levels in bitewing and periapical radiographs for the diagnosis and treatment planning of periodontal disease," and summarized the test as "an average difference of only 0.3mm" from dentist consensus. It also said Overjet's software was "already in daily use by multiple large dental insurance companies to automate and improve the accuracy of dental claims review" and that the company had raised over $10.5 million. The FDA record makes no claim about firsts; that phrase is the company's.

Who it applies to#

Any practice that runs Overjet on its bitewings and periapicals, and the people the indication names: dentists and hygienists, as trained professionals, on a prescription basis. The measurement the FDA cleared is per tooth, mesial and distal, on those two image types; the summary describes it as an aid to the diagnostic process, and the indication says it may not be the sole basis for a diagnosis.

Pediatric and mixed-dentition patients are outside it. The population is 22 and older with no remaining primary teeth, and Overjet told the FDA it had not evaluated performance on primary dentition. Overjet later obtained a separate pediatric clearance for caries (K233738, March 4, 2024, per the FDA database), which does not change the bone-level indication described here.

Practices that use the output for insurance documentation should note what the release, not the FDA, says: payer-side use of Overjet's software in claims review predates this clearance. The clearance governs what the software may claim to do for the clinician.

What changed#

Overjet called this the first FDA clearance to bring AI into the dental practice; the FDA's own route was ordinary. Dental Assist was cleared by substantial equivalence to an existing orthodontic image program under the long-standing 892.2050 image-processing regulation, not through a new category for artificial intelligence. What it established was a cleared, written indication for an AI measurement of bone levels that a practice can hold a vendor to.

The product line then widened through separate 510(k)s. Overjet Caries Assist was cleared May 10, 2022 as K212519 under 21 CFR 892.2070 (medical image analyzer, product code MYN), with a summary describing software "intended to aid in the detection and segmentation of caries on bitewing radiographs." After that, the FDA's 510(k) database lists Overjet clearances for Calculus Assist (December 16, 2022), Caries Assist (March 27, 2023), Periapical Radiolucency Assist (September 21, 2023), Charting Assist (February 23 and August 27, 2024), Caries Assist-Pediatric (March 4, 2024), Image Enhancement Assist (September 9, 2024), CBCT Assist (December 5, 2025), the Iris Intelligent Imaging System (April 10, 2026) and Multi-Image Caries and Charting Assist (August 6, 2026). Each carries its own indication; none of them expands what K210187 says.

Where the source is#

The FDA's 510(k) database record for K210187 is the source page; its PDF holds the clearance letter, the Indications for Use form (Form FDA 3881) and Overjet's 510(k) summary, with the pages after the letter stored as scanned images rather than searchable text. The K212519 record and PDF document the caries clearance the same way, and 21 CFR 892.2050 on eCFR is the regulation. Overjet's release is on PR Newswire. FDA documents are public domain; the openFDA API, which returns the same database records, carries the FDA's note that results should be treated as unvalidated.