Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-0798
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-08-17T06:21:28.314392Z
status: published
lens: g_space
sector: health
headline: Validation of artificial intelligence-assisted CBCT analysis for predicting inferior alveolar nerve proximity to impacted mandibular third molars: a diagnostic accuracy study
dek: This study aimed to evaluate the diagnostic accuracy of an artificial intelligence (AI)-assisted cone-beam computed tomography (CBCT) analysis system for predicting the spatial proximity of the inferior alveolar nerve (IAN) to impacted mandibular third molars (M3M), using expert radiologist assessment as the reference standard. A retrospective diagnostic accuracy study was conducted on an internal institutional cohort of 312 patients (mean age 28.21 ± 6.62 years; January 2021-December 2024). A deep learning syst…
gain_title: An AI system using modified U-Net segmentation automatically classified inferior alveolar nerve proximity to impacted mandibular third molars on CBCT with 90.1% accuracy and reduced analysis time from ~189 seconds to ~4.8 seconds compared to expert radiologists.
problem_title: (none)
trace_subject: (none)
gain_reading: An AI system using modified U-Net segmentation automatically classified inferior alveolar nerve proximity to impacted mandibular third molars on CBCT with 90.1% accuracy and reduced analysis time from ~189 seconds to ~4.8 seconds compared to expert radiologists.
gain_evidence: The AI system achieved an overall accuracy of 90.1% (438/486; 95% CI: 87.1-92.5%)
problem_reading: (none)
problem_evidence: (none)
quick_read: Researchers retrospectively tested a deep learning system that automatically segments the inferior alveolar nerve canal and impacted mandibular third molars on CBCT and classifies their spatial relationship. On an independent hold-out set of 486 sites, the system reached 90.1% overall accuracy and 0.925 weighted AUC against two senior radiologists, with processing time of 4.75 seconds versus 189.12 seconds for experts.

The result matters because IAN injury risk assessment is a key preoperative step for wisdom tooth removal, and automated CBCT analysis could support faster decision-making if validated. Uncertainty remains because the reference was radiographic interpretation, not intraoperative nerve exposure or postoperative neurosensory outcomes, and the data were single-center retrospective, requiring prospective multicenter study.
limitation: Reference standard was radiographic expert assessment rather than intraoperative findings, and authors state prospective multicenter validation with surgical outcomes is needed before clinical deployment.
tag: Evidence-backed gain
key_points: Retrospective diagnostic accuracy study on 312 patients (Jan 2021-Dec 2024) with independent hold-out test set of 486 M3 sites from 283 patients. | Deep learning system automatically segmented IAN canal and M3M and classified relationship as no contact (>2 mm), proximity (0-2 mm), and contact/overlap. | Reference standard was two senior oral and maxillofacial radiologists; metrics included sensitivity, specificity, PPV, NPV, AUC, Cohen's kappa with Wilson and bootstrap CIs. | Reported weighted AUC 0.925, Cohen's ba 0.851, mean DSC 0.90 for IAN canal and 0.93 for M3M, and Bland-Altman mean difference 0.06 mm.
rundown: The study used an internal institutional cohort of 312 patients (mean age 28.21 7 6.62 years) and evaluated performance on 486 M3 sites. The system segmented the IAN canal and impacted M3M and assigned three proximity categories, with per-category sensitivity 88.2% to 91.2% and specificity 92.8% to 97.4%.

Subgroup analysis found highest accuracy for mesioangular (93.1%) and horizontal (91.6%) impaction. Segmentation quality was reported as mean DSC 0.90 7 0.04 for IAN canal and 0.93 7 0.03 for M3M, with Bland-Altman mean difference 0.06 mm (95% LoA: -0.63 to 0.75 mm).
sources:
- peer_reviewed | BMC Oral Health | https://doi.org/10.1186/s12903-026-08690-z | 2026-08-15
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

b99b1ee9ffcb25abe01dcd165998362d7661110ade6a4eb4fe8b3f4ef641ce46

previous

0000000000000000000000000000000000000000000000000000000000000000