The influence of a deep learning tool on the performance of oral and maxillofacial radiologists in the detection of apical radiolucencies
Bibliographic record
Abstract
OBJECTIVES: This study aimed to assess the impact of a deep learning model on oral radiologists' ability to detect periapical radiolucencies on periapical radiographs. The secondary objective was to conduct a regression analysis to evaluate the effects of years of experience, time to diagnose, and specialty. METHODS: This study used an annotated dataset and a beta version of a deep learning model (Denti.AI). The testing subset comprised 68 intraoral periapical radiographs confirmed with cone-beam computed tomography for the presence/absence of apical radiolucencies. Four oral radiologists participated in a cross-over reading scenario, analysing the radiographs under 2 conditions: initially without AI assistance and later with AI predictions. The study evaluated reader performance using Alternative Free-Response Receiver Operating Characteristic - Area Under the Curve (AFROC-AUC), sensitivity, specificity, and Receiver Operating Characteristic-Area Under the Curve (ROC-AUC) per case. It also assessed sensitivity per lesion. Regression analysis investigated how experience, time spent on images, and specialty influenced reader performance. RESULTS: No statistically significant differences were found in AFROC-AUC, sensitivity, specificity, and ROC-AUC. Regression analysis identified factors influencing diagnostic outcomes: unaided reading significantly prolonged diagnostic time (β = 12, 95% CI, 11-13, P < 0.001), while radiologists' professional status was positively associated with diagnostic accuracy (β = 0.02, 95% CI, 0.00-0.04, P = 0.015). These findings underscore the impact of AI on diagnostic efficiency and the critical role of radiologists' experience in diagnostic accuracy. CONCLUSION: AI did not significantly enhance radiologists' overall diagnostic accuracy. However, it showed potential to enhance efficiency, particularly advantageous for non-expert clinicians. The expertise of radiologists remains vital for accuracy, underscoring the complementary role of AI in dental diagnostics.
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.
How this classification was reachedexpand
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) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.002 |
| 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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".