Radiomics-based classification of pediatric dental trauma in periapical radiographs: a preliminary study

Abstract

Purpose: To evaluate the feasibility and effectiveness of radiomic analysis of the periapical region on periapical radiographs for classifying the subtypes of dental trauma in pediatric patients.

Methods: A retrospective analysis was conducted on 111 pediatric patients who presented with dental trauma and underwent periapical radiography. Patients were categorized into tooth concussion (n = 23) and tooth fracture (n = 88) groups on the basis of the type of injury. Patients were randomly stratified into training (n = 78; concussion: 16, fracture: 62) and testing (n = 33; concussion: 7, fracture: 26) cohorts at a 7:3 ratio. Regions of interest were manually delineated around the apical foramen, and radiomic features were subsequently extracted. Feature selection was performed using the intraclass correlation coefficient, the Pearson correlation coefficient, and one-way ANOVA. A support vector machine classifier was constructed based on the selected features. The performance of the radiomic model was evaluated using the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity.

Results: A total of 21 radiomic features were selected to construct the final model. In the training cohort, the model achieved an AUC of 0.766 (95% confidence interval (CI): 0.605–0.927), with a sensitivity of 1.000 and a specificity of 0.500 in differentiating the subtypes of dental trauma. In the testing cohort, the model yielded an AUC of 0.758 (95% CI: 0.538–0.979), with a sensitivity of 0.808 and a specificity of 0.714.

Conclusion: Radiomic analysis of periapical radiographs shows promise in distinguishing between tooth concussion and fracture in pediatric patients. Further validation is needed to confirm its clinical utility and broader applicability.

Supplementary Information: The online version contains supplementary material available at 10.1186/s12880-025-01877-w.

Keywords: Children; Dental trauma; Periapical radiography; Radiomics.

Fig. 1

Study pipeline used in this research

Fig. 2

Example of region of interest segmentation

Fig. 3

Intraclass correlation coefficient distribution of each radiomic feature extracted from the two sets of segmentations

Fig. 4

The optimal radiomic features ranked according to the F values of ANOVA

Fig. 5

Comparison of the top five radiomic features between groups in the entire cohort. The label “0” refers to tooth concussion, and the label “1” refers to tooth fracture

Fig. 6

Receiver operating characteristic curves of the radiomic model without (A) and with (B) Synthetic Minority Oversampling Technique in the training and testing cohorts

Fig. 7

Calibration curves (A) and decision curve (B) of the radiomic model without using the Synthetic Minority Oversampling Technique. The red curve in decision curve analysis (B) demonstrates the net benefit across varying high-risk thresholds. The model offers greater clinical utility compared to treating all or none, particularly within the 0.10–0.90 range

Fig. 8

A case of tooth concussion. The periapical radiograph shows a low-density area around the apical region, widening of the periodontal ligament space, and apical foramen development at Nolla stages 7–8