The Accuracy of Algorithms Used by Artificial Intelligence in Cephalometric Points Detection: A Systematic Review
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
The use of artificial intelligence in orthodontics is emerging as a tool for localizing cephalometric points in two-dimensional X-rays. AI systems are being evaluated for their accuracy and efficiency compared to conventional methods performed by professionals. The main objective of this study is to identify the artificial intelligence algorithms that yield the best results for cephalometric landmark localization, along with their learning system. A literature search was conducted across PubMed-MEDLINE, Cochrane, Scopus, IEEE Xplore, and Web of Science. Observational and experimental studies from 2013 to 2023 assessing the detection of at least 13 cephalometric landmarks in two-dimensional radiographs were included. Studies requiring advanced computer engineering knowledge or involving patients with anomalies, syndromes, or orthodontic appliances, were excluded. Risk of bias was assessed using Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) and Newcastle-Ottawa Scale (NOS) tools. Of 385 references, 13 studies met the inclusion criteria (1 diagnostic accuracy study and 12 retrospective cohorts). Six were high-risk, and seven were low-risk. Convolutional neural networks (CNN)-based AI algorithms showed point localization accuracy ranging from 64.3 to 97.3%, with a mean error of 1.04 mm ± 0.89 to 3.40 mm ± 1.57, within the clinical range of 2 mm. YOLOv3 demonstrated improvements over its earlier version. CNN have proven to be the most effective AI system for detecting cephalometric points in radiographic images. Although CNN-based algorithms generate results very quickly and reproducibly, they still do not achieve the accuracy of orthodontists.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.002 | 0.001 |
| Bibliometrics | 0.001 | 0.005 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it