Regression-based correction of panoramic radiographic distortion for tooth/root length estimation: a CBCT-referenced model-development study

Abstract

Accurate radiographic tooth/root length estimation can support endodontic treatment planning, but panoramic radiographs are affected by magnification and regional distortion. This retrospective model-development study evaluated the accuracy of panoramic radiographic measurements against cone-beam computed tomography (CBCT)-derived reference measurements and developed regression-based correction equations for estimating tooth/root length from panoramic measurements. The study included 156 patients with paired orthopantomogram (OPG) and CBCT records. A balanced tooth/root-level dataset of 1,300 paired observations was created, with 100 observations in each of 13 tooth/root categories. CBCT-derived tooth/root lengths were used as the reference measurements. Two calibrated investigators performed repeated measurements, and intra- and inter-examiner reliability were assessed using intraclass correlation coefficients. Paired comparisons, linear regression modelling, mean absolute error, root mean square error, clinical threshold analysis, and Bland-Altman agreement analysis were performed. The overall mean OPG measurement was 20.33 ± 3.04 mm, whereas the mean CBCT-derived reference measurement was 21.63 ± 2.75 mm. Uncorrected OPG measurements significantly underestimated tooth/root length by -1.30 mm. The overall regression model showed a positive relationship between OPG and CBCT-derived measurements and explained 48.9% of the variance. Regression-based correction reduced the apparent mean absolute error within the present dataset from 2.00 mm to 1.56 mm using the overall equation. For the category-specific equations used in the desktop application, 53.6% of corrected predictions were within ± 1.0 mm and 83.8% were within ± 2.0 mm of the CBCT-derived reference measurements. Panoramic radiographs generally underestimated CBCT-derived tooth/root length, with distortion varying by tooth/root category. The correction model reduced apparent measurement error within the study dataset, but no independent external validation was performed. Therefore, the equations should be regarded as exploratory preoperative estimation tools pending patient-level internal validation and independent external validation. Regression-based correction of panoramic radiographic measurements may improve preoperative estimation of tooth/root length, particularly in regions with greater panoramic distortion. The model may support endodontic planning, but it should be used only as an adjunctive radiographic support tool and not as a replacement for established clinical working-length determination methods.