Measurements and landmarks

Proposal methods, sections 3.8–3.9 (pp. 18–23).

Use measurements made by the radiologist in the study’s panoramic and CBCT workstations. Enter distances in millimetres.

A

Root apex of the impacted mandibular third molar on OPG.

B

Superior border of the inferior alveolar canal on OPG.

C

Inferior border of the canal on OPG.

D

Inferior mandibular border in the third molar region on OPG.

E

Root apex on the CBCT cross-section at closest proximity.

F

Adjacent superior canal border on that CBCT cross-section.

Use the same vertical reference path for the OPG distances, following the workstation protocol. For EF, select the closest buccolingual CBCT slice in the multiplanar reconstruction. Apply a consistent rule for the apex when more than one root is visible.

Ratios and proximity bands

EF exactly 2 mm is assigned to the high band. The app classifies using full calculation precision before rounding for display.

AD cancels out of the final index, which simplifies to 100 × BC / (AB + AC). With collinear landmarks, a projected apex between B and C yields an NRI of 100 across the canal. Check this potential saturation in the cohort.

Records where BC exceeds AB + AC by more than the 0.015 mm tolerance for rounding to 0.01 mm are flagged for review and retained. Check landmark alignment and transcription before analysis.

Cohort and data handling

The proposal targets 199 patients aged 20–59 from the Al-Karama Specialized Dental Center, Baghdad, with eligible paired OPG and CBCT records from 2020–2025.

• Screen for medical eligibility, fully formed roots, suitable image quality and complete records. Apply the proposal exclusions for prior mandibular surgery, relevant pathology and craniofacial anomalies.
• Use one study code per patient. Pair both modalities for each impacted FDI tooth, 38 or 48. Age summaries count each patient once.
• Enter coded records on a password-protected study computer. Browser storage stays on this device and browser profile. Export CSV backups to the approved study location.
• Use the optional EF readings for the same fixed reader and session roles across the repeatability subset.

Analysis and interpretation

• Use the three-category table and unweighted Cohen’s κ for agreement. Spearman’s ρ reports association with both ordinal CBCT band and continuous EF.
• Binary accuracy uses the selected pair of fixed thresholds. ROC AUC uses continuous NRI and the selected CBCT reference cutoff. Predictive values describe this retrospective cohort. Prespecify the primary binary cutoff before evaluating study outcomes.
• All-teeth analysis resamples patients with their teeth together for 95% percentile intervals (1,200 draws). An interval requires at least 4 patients and 90% defined replicates.
• Use Stuart–Maxwell for three paired categories and exact McNemar after binary thresholding. Their p-values require one tooth per patient. Prespecify the preferred side before evaluating outcomes.
• Repeatability uses single-reading, absolute-agreement, two-way mixed-effects ICC and three-band Cohen’s κ for fixed readers and sessions. The displayed estimates are descriptive.
• The screening scenario flags a patient if any recorded tooth screens positive. The reference is radiographic proximity. Observed nerve injury, actual clinical indications for CBCT, dose and cost would require additional data.

Stuart–Maxwell testPatient bootstrapICC specification