Cone-beam computed tomography radiomics for differentiating clinicoradiologic labels of osteosclerosis,...

To develop and internally validate cone-beam computed tomography radiomics for recorded clinicoradiologic classification of osteosclerosis, cemento-osseous dysplasia and odontoma, with secondary same-case three-dimensional convolutional neural-network benchmarking.

This retrospective single-centre study included 126 lesions from 126 patients. From manual three-dimensional segmentations, 107 original radiomics features were extracted. A class-weighted least absolute shrinkage and selection operator one-vs-rest logistic-regression model was evaluated using 100 repeats of stratified five-fold held-out resampling. No independent histopathology report was retrievable in the retained study archive. A secondary strict 124-lesion three-dimensional analysis used 0.30-mm isotropic variable-size lesion-centred inputs, at least 5 mm of physical context and nested architecture/epoch selection; locked radiomics was evaluated on the same cases and outer folds.

Primary radiomics achieved accuracy 0.825, balanced accuracy 0.842, macro F1-score 0.834 and macro area under the curve 0.932. In the matched 124-lesion analysis, macro area under the curve was 0.935 (95% bias-corrected and accelerated confidence interval 0.889-0.963) for radiomics and 0.601 (0.516-0.682) for the nested three-dimensional network. The network showed marked odontoma-class collapse (sensitivity 1/22, 0.045).

Radiomics showed promising internal discrimination of the recorded labels, whereas the secondary three-dimensional benchmark was unstable for odontoma in this small cohort. These findings do not establish diagnostic accuracy against histopathology or general algorithmic superiority; external validation, standardised imaging and stronger reference standards are required.

This study extends cone-beam computed tomography radiomics to radiopaque/mineralised jaw-lesion recorded labels and reports a same-case nested three-dimensional benchmark with explicit disclosure of class-specific failure.