Periapical and Panoramic Radiographs Yield Non-comparable Quantitative Texture Metrics Despite Physical ROI...

To determine whether quantitative trabecular texture metrics derived from periapical and panoramic radiographs of the same tooth yield numerically comparable values, and to verify the technical implementation fidelity of a semi-automated analysis tool against the established ImageJ/FracLac pipeline.

In this retrospective, single-center paired study, 54 matched panoramic-periapical image pairs (108 ROIs) from the same tooth were analyzed. ROI size was physically standardized to a 3.08-mm edge length starting 1 mm apical to the root apex. Fractal dimension (FD), lacunarity, GLCM contrast, and entropy were computed with a semi-automated tool implementing the ImageJ/FracLac pipeline. Normality of paired differences was assessed with the Shapiro-Wilk test; paired t-test, or Wilcoxon signed-rank test was applied with Holm-adjusted p-values. Robustness was probed by resampling both modalities to a common pixel grid and by equalizing the box-counting scale set. Implementation fidelity was evaluated using ICC(A,1), Bland-Altman analysis, and Lin's concordance correlation coefficient.

Periapical entropy was significantly higher than panoramic (mean difference + 1.05; 95% CI 0.80-1.31; Cohen's dz = 1.12; Holm-adjusted p < 0.001) and remained significant after resampling both modalities to a common pixel grid (+ 0.96; p < 0.001), indicating a modality effect not attributable to resolution alone. Fractal dimension, lacunarity, and contrast showed no significant modality difference (all Holm-adjusted p = 1.00); fractal dimension in particular varied with ROI size and processing choices, indicating limited robustness under the conditions tested. The semi-automated tool reproduced the ImageJ pipeline with near-zero bias (ICC(A,1) ≥ 0.998). Intra-observer retest ICC ranged from 0.853 to 0.962.

In this retrospective, single-center paired study, periapical and panoramic radiographs of the same tooth yielded systematically different entropy values-a difference that persisted after resampling both modalities to a common pixel grid and therefore reflects a genuine modality effect rather than a resolution artifact. In contrast, fractal dimension, lacunarity, and GLCM contrast did not differ significantly under the predefined analytical conditions. The systematic cross-modality difference established here is thus specific to entropy and concerns the numerical comparability of this metric rather than diagnostic equivalence; maintaining the same imaging modality is recommended for longitudinal quantitative entropy follow-up to avoid systematic measurement bias. The semi-automated tool reproduced the reference pipeline within a streamlined workflow.