orthodontic.ai

Cases · did it work?

Did it work? Case notes

Each case keeps three things apart: what vendors or sponsors claim, what we could check ourselves, and the limits. Every source is linked and marked as read or not. Sorted by strongest evidence.

  1. L3 · Peer-reviewed study cephalometrics · imaging · FDA 510(k)

    AI cephalometric landmarking: inside 2 mm, most of the time

    A 2021 meta-analysis found deep-learning tools placed about 80% of landmarks within 2 mm. Two FDA-cleared AI cephalometric software products report passing their own bench thresholds. The studies behind both carry real bias caveats.

    Short answer Mostly, as a first draft. Accuracy is near the 2 mm threshold clinicians commonly accept, with high risk of bias in the underlying studies and every tracing still needing review.

    Updated 2026-10-31

  2. L2 · Regulatory record aligners · treatment planning · FDA 510(k)

    AI in aligner treatment planning: what an FDA clearance does and does not tell you

    FDA cleared an optional AI-enabled prediction service inside a clear-aligner treatment-planning system on Sep 25, 2026. We read the clearance summary and set out what it does and does not show.

    Short answer Too early to say. The clearance summary reports that pre-set acceptance criteria were met but publishes no figures.

    Updated 2026-10-31

  3. L2 · Regulatory record remote monitoring · aligners · FDA De Novo

    Remote monitoring between visits: fewer appointments, same finish?

    An FDA-cleared remote-monitoring app was granted De Novo classification in May 2024. We read the FDA decision summary and set out what it supports.

    Short answer Not shown by the sources in this draft. The FDA summary describes the app as an aid, not a stand-alone decision tool, and reports the sponsor's own detection figures. It does not measure visits, cost or outcomes.

    Updated 2026-10-31

The evidence ladder

No case here reaches level 4: we did not find a randomized trial for any of these three topics in this pass.

  1. 1 Vendor claim Press release, website or brochure from the company selling the tool. Tells you what to ask, not what is true.
  2. 2 Regulatory record FDA 510(k) or De Novo record and summary. Describes the sponsor's own testing as summarized for FDA. Clearance is not a comparison or an outcome study.
  3. 3 Peer-reviewed study Published cohort, accuracy study or systematic review of non-randomized studies. Has a method section; bias and setting limits still apply.
  4. 4 Randomized trial Patients or clinicians randomly assigned to use the tool or not, with an independent reference standard. We have not yet found one for these topics.